git.oblomov.eu Git - linux-2.6/blob - drivers/net/s2io.c

   1 /************************************************************************
   2  * s2io.c: A Linux PCI-X Ethernet driver for Neterion 10GbE Server NIC
   3  * Copyright(c) 2002-2005 Neterion Inc.
   4
   5  * This software may be used and distributed according to the terms of
   6  * the GNU General Public License (GPL), incorporated herein by reference.
   7  * Drivers based on or derived from this code fall under the GPL and must
   8  * retain the authorship, copyright and license notice.  This file is not
   9  * a complete program and may only be used when the entire operating
  10  * system is licensed under the GPL.
  11  * See the file COPYING in this distribution for more information.
  12  *
  13  * Credits:
  14  * Jeff Garzik          : For pointing out the improper error condition
  15  *                        check in the s2io_xmit routine and also some
  16  *                        issues in the Tx watch dog function. Also for
  17  *                        patiently answering all those innumerable
  18  *                        questions regaring the 2.6 porting issues.
  19  * Stephen Hemminger    : Providing proper 2.6 porting mechanism for some
  20  *                        macros available only in 2.6 Kernel.
  21  * Francois Romieu      : For pointing out all code part that were
  22  *                        deprecated and also styling related comments.
  23  * Grant Grundler       : For helping me get rid of some Architecture
  24  *                        dependent code.
  25  * Christopher Hellwig  : Some more 2.6 specific issues in the driver.
  26  *
  27  * The module loadable parameters that are supported by the driver and a brief
  28  * explaination of all the variables.
  29  * rx_ring_num : This can be used to program the number of receive rings used
  30  * in the driver.
  31  * rx_ring_sz: This defines the number of descriptors each ring can have. This
  32  * is also an array of size 8.
  33  * rx_ring_mode: This defines the operation mode of all 8 rings. The valid
  34  *              values are 1, 2 and 3.
  35  * tx_fifo_num: This defines the number of Tx FIFOs thats used int the driver.
  36  * tx_fifo_len: This too is an array of 8. Each element defines the number of
  37  * Tx descriptors that can be associated with each corresponding FIFO.
  38  ************************************************************************/
  39
  40 #include <linux/config.h>
  41 #include <linux/module.h>
  42 #include <linux/types.h>
  43 #include <linux/errno.h>
  44 #include <linux/ioport.h>
  45 #include <linux/pci.h>
  46 #include <linux/dma-mapping.h>
  47 #include <linux/kernel.h>
  48 #include <linux/netdevice.h>
  49 #include <linux/etherdevice.h>
  50 #include <linux/skbuff.h>
  51 #include <linux/init.h>
  52 #include <linux/delay.h>
  53 #include <linux/stddef.h>
  54 #include <linux/ioctl.h>
  55 #include <linux/timex.h>
  56 #include <linux/sched.h>
  57 #include <linux/ethtool.h>
  58 #include <linux/workqueue.h>
  59 #include <linux/if_vlan.h>
  60 #include <linux/ip.h>
  61 #include <linux/tcp.h>
  62 #include <net/tcp.h>
  63
  64 #include <asm/system.h>
  65 #include <asm/uaccess.h>
  66 #include <asm/io.h>
  67 #include <asm/div64.h>
  68
  69 /* local include */
  70 #include "s2io.h"
  71 #include "s2io-regs.h"
  72
  73 #define DRV_VERSION "2.0.11.2"
  74
  75 /* S2io Driver name & version. */
  76 static char s2io_driver_name[] = "Neterion";
  77 static char s2io_driver_version[] = DRV_VERSION;
  78
  79 static int rxd_size[4] = {32,48,48,64};
  80 static int rxd_count[4] = {127,85,85,63};
  81
  82 static inline int RXD_IS_UP2DT(RxD_t *rxdp)
  83 {
  84         int ret;
  85
  86         ret = ((!(rxdp->Control_1 & RXD_OWN_XENA)) &&
  87                 (GET_RXD_MARKER(rxdp->Control_2) != THE_RXD_MARK));
  88
  89         return ret;
  90 }
  91
  92 /*
  93  * Cards with following subsystem_id have a link state indication
  94  * problem, 600B, 600C, 600D, 640B, 640C and 640D.
  95  * macro below identifies these cards given the subsystem_id.
  96  */
  97 #define CARDS_WITH_FAULTY_LINK_INDICATORS(dev_type, subid) \
  98         (dev_type == XFRAME_I_DEVICE) ?                 \
  99                 ((((subid >= 0x600B) && (subid <= 0x600D)) || \
 100                  ((subid >= 0x640B) && (subid <= 0x640D))) ? 1 : 0) : 0
 101
 102 #define LINK_IS_UP(val64) (!(val64 & (ADAPTER_STATUS_RMAC_REMOTE_FAULT | \
 103                                       ADAPTER_STATUS_RMAC_LOCAL_FAULT)))
 104 #define TASKLET_IN_USE test_and_set_bit(0, (&sp->tasklet_status))
 105 #define PANIC   1
 106 #define LOW     2
 107 static inline int rx_buffer_level(nic_t * sp, int rxb_size, int ring)
 108 {
 109         int level = 0;
 110         mac_info_t *mac_control;
 111
 112         mac_control = &sp->mac_control;
 113         if ((mac_control->rings[ring].pkt_cnt - rxb_size) > 16) {
 114                 level = LOW;
 115                 if (rxb_size <= rxd_count[sp->rxd_mode]) {
 116                         level = PANIC;
 117                 }
 118         }
 119
 120         return level;
 121 }
 122
 123 /* Ethtool related variables and Macros. */
 124 static char s2io_gstrings[][ETH_GSTRING_LEN] = {
 125         "Register test\t(offline)",
 126         "Eeprom test\t(offline)",
 127         "Link test\t(online)",
 128         "RLDRAM test\t(offline)",
 129         "BIST Test\t(offline)"
 130 };
 131
 132 static char ethtool_stats_keys[][ETH_GSTRING_LEN] = {
 133         {"tmac_frms"},
 134         {"tmac_data_octets"},
 135         {"tmac_drop_frms"},
 136         {"tmac_mcst_frms"},
 137         {"tmac_bcst_frms"},
 138         {"tmac_pause_ctrl_frms"},
 139         {"tmac_any_err_frms"},
 140         {"tmac_vld_ip_octets"},
 141         {"tmac_vld_ip"},
 142         {"tmac_drop_ip"},
 143         {"tmac_icmp"},
 144         {"tmac_rst_tcp"},
 145         {"tmac_tcp"},
 146         {"tmac_udp"},
 147         {"rmac_vld_frms"},
 148         {"rmac_data_octets"},
 149         {"rmac_fcs_err_frms"},
 150         {"rmac_drop_frms"},
 151         {"rmac_vld_mcst_frms"},
 152         {"rmac_vld_bcst_frms"},
 153         {"rmac_in_rng_len_err_frms"},
 154         {"rmac_long_frms"},
 155         {"rmac_pause_ctrl_frms"},
 156         {"rmac_discarded_frms"},
 157         {"rmac_usized_frms"},
 158         {"rmac_osized_frms"},
 159         {"rmac_frag_frms"},
 160         {"rmac_jabber_frms"},
 161         {"rmac_ip"},
 162         {"rmac_ip_octets"},
 163         {"rmac_hdr_err_ip"},
 164         {"rmac_drop_ip"},
 165         {"rmac_icmp"},
 166         {"rmac_tcp"},
 167         {"rmac_udp"},
 168         {"rmac_err_drp_udp"},
 169         {"rmac_pause_cnt"},
 170         {"rmac_accepted_ip"},
 171         {"rmac_err_tcp"},
 172         {"\n DRIVER STATISTICS"},
 173         {"single_bit_ecc_errs"},
 174         {"double_bit_ecc_errs"},
 175         ("lro_aggregated_pkts"),
 176         ("lro_flush_both_count"),
 177         ("lro_out_of_sequence_pkts"),
 178         ("lro_flush_due_to_max_pkts"),
 179         ("lro_avg_aggr_pkts"),
 180 };
 181
 182 #define S2IO_STAT_LEN sizeof(ethtool_stats_keys)/ ETH_GSTRING_LEN
 183 #define S2IO_STAT_STRINGS_LEN S2IO_STAT_LEN * ETH_GSTRING_LEN
 184
 185 #define S2IO_TEST_LEN   sizeof(s2io_gstrings) / ETH_GSTRING_LEN
 186 #define S2IO_STRINGS_LEN        S2IO_TEST_LEN * ETH_GSTRING_LEN
 187
 188 #define S2IO_TIMER_CONF(timer, handle, arg, exp)                \
 189                         init_timer(&timer);                     \
 190                         timer.function = handle;                \
 191                         timer.data = (unsigned long) arg;       \
 192                         mod_timer(&timer, (jiffies + exp))      \
 193
 194 /* Add the vlan */
 195 static void s2io_vlan_rx_register(struct net_device *dev,
 196                                         struct vlan_group *grp)
 197 {
 198         nic_t *nic = dev->priv;
 199         unsigned long flags;
 200
 201         spin_lock_irqsave(&nic->tx_lock, flags);
 202         nic->vlgrp = grp;
 203         spin_unlock_irqrestore(&nic->tx_lock, flags);
 204 }
 205
 206 /* Unregister the vlan */
 207 static void s2io_vlan_rx_kill_vid(struct net_device *dev, unsigned long vid)
 208 {
 209         nic_t *nic = dev->priv;
 210         unsigned long flags;
 211
 212         spin_lock_irqsave(&nic->tx_lock, flags);
 213         if (nic->vlgrp)
 214                 nic->vlgrp->vlan_devices[vid] = NULL;
 215         spin_unlock_irqrestore(&nic->tx_lock, flags);
 216 }
 217
 218 /*
 219  * Constants to be programmed into the Xena's registers, to configure
 220  * the XAUI.
 221  */
 222
 223 #define SWITCH_SIGN     0xA5A5A5A5A5A5A5A5ULL
 224 #define END_SIGN        0x0
 225
 226 static const u64 herc_act_dtx_cfg[] = {
 227         /* Set address */
 228         0x8000051536750000ULL, 0x80000515367500E0ULL,
 229         /* Write data */
 230         0x8000051536750004ULL, 0x80000515367500E4ULL,
 231         /* Set address */
 232         0x80010515003F0000ULL, 0x80010515003F00E0ULL,
 233         /* Write data */
 234         0x80010515003F0004ULL, 0x80010515003F00E4ULL,
 235         /* Set address */
 236         0x801205150D440000ULL, 0x801205150D4400E0ULL,
 237         /* Write data */
 238         0x801205150D440004ULL, 0x801205150D4400E4ULL,
 239         /* Set address */
 240         0x80020515F2100000ULL, 0x80020515F21000E0ULL,
 241         /* Write data */
 242         0x80020515F2100004ULL, 0x80020515F21000E4ULL,
 243         /* Done */
 244         END_SIGN
 245 };
 246
 247 static const u64 xena_mdio_cfg[] = {
 248         /* Reset PMA PLL */
 249         0xC001010000000000ULL, 0xC0010100000000E0ULL,
 250         0xC0010100008000E4ULL,
 251         /* Remove Reset from PMA PLL */
 252         0xC001010000000000ULL, 0xC0010100000000E0ULL,
 253         0xC0010100000000E4ULL,
 254         END_SIGN
 255 };
 256
 257 static const u64 xena_dtx_cfg[] = {
 258         0x8000051500000000ULL, 0x80000515000000E0ULL,
 259         0x80000515D93500E4ULL, 0x8001051500000000ULL,
 260         0x80010515000000E0ULL, 0x80010515001E00E4ULL,
 261         0x8002051500000000ULL, 0x80020515000000E0ULL,
 262         0x80020515F21000E4ULL,
 263         /* Set PADLOOPBACKN */
 264         0x8002051500000000ULL, 0x80020515000000E0ULL,
 265         0x80020515B20000E4ULL, 0x8003051500000000ULL,
 266         0x80030515000000E0ULL, 0x80030515B20000E4ULL,
 267         0x8004051500000000ULL, 0x80040515000000E0ULL,
 268         0x80040515B20000E4ULL, 0x8005051500000000ULL,
 269         0x80050515000000E0ULL, 0x80050515B20000E4ULL,
 270         SWITCH_SIGN,
 271         /* Remove PADLOOPBACKN */
 272         0x8002051500000000ULL, 0x80020515000000E0ULL,
 273         0x80020515F20000E4ULL, 0x8003051500000000ULL,
 274         0x80030515000000E0ULL, 0x80030515F20000E4ULL,
 275         0x8004051500000000ULL, 0x80040515000000E0ULL,
 276         0x80040515F20000E4ULL, 0x8005051500000000ULL,
 277         0x80050515000000E0ULL, 0x80050515F20000E4ULL,
 278         END_SIGN
 279 };
 280
 281 /*
 282  * Constants for Fixing the MacAddress problem seen mostly on
 283  * Alpha machines.
 284  */
 285 static const u64 fix_mac[] = {
 286         0x0060000000000000ULL, 0x0060600000000000ULL,
 287         0x0040600000000000ULL, 0x0000600000000000ULL,
 288         0x0020600000000000ULL, 0x0060600000000000ULL,
 289         0x0020600000000000ULL, 0x0060600000000000ULL,
 290         0x0020600000000000ULL, 0x0060600000000000ULL,
 291         0x0020600000000000ULL, 0x0060600000000000ULL,
 292         0x0020600000000000ULL, 0x0060600000000000ULL,
 293         0x0020600000000000ULL, 0x0060600000000000ULL,
 294         0x0020600000000000ULL, 0x0060600000000000ULL,
 295         0x0020600000000000ULL, 0x0060600000000000ULL,
 296         0x0020600000000000ULL, 0x0060600000000000ULL,
 297         0x0020600000000000ULL, 0x0060600000000000ULL,
 298         0x0020600000000000ULL, 0x0000600000000000ULL,
 299         0x0040600000000000ULL, 0x0060600000000000ULL,
 300         END_SIGN
 301 };
 302
 303 /* Module Loadable parameters. */
 304 static unsigned int tx_fifo_num = 1;
 305 static unsigned int tx_fifo_len[MAX_TX_FIFOS] =
 306     {[0 ...(MAX_TX_FIFOS - 1)] = 0 };
 307 static unsigned int rx_ring_num = 1;
 308 static unsigned int rx_ring_sz[MAX_RX_RINGS] =
 309     {[0 ...(MAX_RX_RINGS - 1)] = 0 };
 310 static unsigned int rts_frm_len[MAX_RX_RINGS] =
 311     {[0 ...(MAX_RX_RINGS - 1)] = 0 };
 312 static unsigned int rx_ring_mode = 1;
 313 static unsigned int use_continuous_tx_intrs = 1;
 314 static unsigned int rmac_pause_time = 65535;
 315 static unsigned int mc_pause_threshold_q0q3 = 187;
 316 static unsigned int mc_pause_threshold_q4q7 = 187;
 317 static unsigned int shared_splits;
 318 static unsigned int tmac_util_period = 5;
 319 static unsigned int rmac_util_period = 5;
 320 static unsigned int bimodal = 0;
 321 static unsigned int l3l4hdr_size = 128;
 322 #ifndef CONFIG_S2IO_NAPI
 323 static unsigned int indicate_max_pkts;
 324 #endif
 325 /* Frequency of Rx desc syncs expressed as power of 2 */
 326 static unsigned int rxsync_frequency = 3;
 327 /* Interrupt type. Values can be 0(INTA), 1(MSI), 2(MSI_X) */
 328 static unsigned int intr_type = 0;
 329 /* Large receive offload feature */
 330 static unsigned int lro = 0;
 331 /* Max pkts to be aggregated by LRO at one time. If not specified,
 332  * aggregation happens until we hit max IP pkt size(64K)
 333  */
 334 static unsigned int lro_max_pkts = 0xFFFF;
 335
 336 /*
 337  * S2IO device table.
 338  * This table lists all the devices that this driver supports.
 339  */
 340 static struct pci_device_id s2io_tbl[] __devinitdata = {
 341         {PCI_VENDOR_ID_S2IO, PCI_DEVICE_ID_S2IO_WIN,
 342          PCI_ANY_ID, PCI_ANY_ID},
 343         {PCI_VENDOR_ID_S2IO, PCI_DEVICE_ID_S2IO_UNI,
 344          PCI_ANY_ID, PCI_ANY_ID},
 345         {PCI_VENDOR_ID_S2IO, PCI_DEVICE_ID_HERC_WIN,
 346          PCI_ANY_ID, PCI_ANY_ID},
 347         {PCI_VENDOR_ID_S2IO, PCI_DEVICE_ID_HERC_UNI,
 348          PCI_ANY_ID, PCI_ANY_ID},
 349         {0,}
 350 };
 351
 352 MODULE_DEVICE_TABLE(pci, s2io_tbl);
 353
 354 static struct pci_driver s2io_driver = {
 355       .name = "S2IO",
 356       .id_table = s2io_tbl,
 357       .probe = s2io_init_nic,
 358       .remove = __devexit_p(s2io_rem_nic),
 359 };
 360
 361 /* A simplifier macro used both by init and free shared_mem Fns(). */
 362 #define TXD_MEM_PAGE_CNT(len, per_each) ((len+per_each - 1) / per_each)
 363
 364 /**
 365  * init_shared_mem - Allocation and Initialization of Memory
 366  * @nic: Device private variable.
 367  * Description: The function allocates all the memory areas shared
 368  * between the NIC and the driver. This includes Tx descriptors,
 369  * Rx descriptors and the statistics block.
 370  */
 371
 372 static int init_shared_mem(struct s2io_nic *nic)
 373 {
 374         u32 size;
 375         void *tmp_v_addr, *tmp_v_addr_next;
 376         dma_addr_t tmp_p_addr, tmp_p_addr_next;
 377         RxD_block_t *pre_rxd_blk = NULL;
 378         int i, j, blk_cnt, rx_sz, tx_sz;
 379         int lst_size, lst_per_page;
 380         struct net_device *dev = nic->dev;
 381         unsigned long tmp;
 382         buffAdd_t *ba;
 383
 384         mac_info_t *mac_control;
 385         struct config_param *config;
 386
 387         mac_control = &nic->mac_control;
 388         config = &nic->config;
 389
 390
 391         /* Allocation and initialization of TXDLs in FIOFs */
 392         size = 0;
 393         for (i = 0; i < config->tx_fifo_num; i++) {
 394                 size += config->tx_cfg[i].fifo_len;
 395         }
 396         if (size > MAX_AVAILABLE_TXDS) {
 397                 DBG_PRINT(ERR_DBG, "%s: Requested TxDs too high, ",
 398                           __FUNCTION__);
 399                 DBG_PRINT(ERR_DBG, "Requested: %d, max supported: 8192\n", size);
 400                 return FAILURE;
 401         }
 402
 403         lst_size = (sizeof(TxD_t) * config->max_txds);
 404         tx_sz = lst_size * size;
 405         lst_per_page = PAGE_SIZE / lst_size;
 406
 407         for (i = 0; i < config->tx_fifo_num; i++) {
 408                 int fifo_len = config->tx_cfg[i].fifo_len;
 409                 int list_holder_size = fifo_len * sizeof(list_info_hold_t);
 410                 mac_control->fifos[i].list_info = kmalloc(list_holder_size,
 411                                                           GFP_KERNEL);
 412                 if (!mac_control->fifos[i].list_info) {
 413                         DBG_PRINT(ERR_DBG,
 414                                   "Malloc failed for list_info\n");
 415                         return -ENOMEM;
 416                 }
 417                 memset(mac_control->fifos[i].list_info, 0, list_holder_size);
 418         }
 419         for (i = 0; i < config->tx_fifo_num; i++) {
 420                 int page_num = TXD_MEM_PAGE_CNT(config->tx_cfg[i].fifo_len,
 421                                                 lst_per_page);
 422                 mac_control->fifos[i].tx_curr_put_info.offset = 0;
 423                 mac_control->fifos[i].tx_curr_put_info.fifo_len =
 424                     config->tx_cfg[i].fifo_len - 1;
 425                 mac_control->fifos[i].tx_curr_get_info.offset = 0;
 426                 mac_control->fifos[i].tx_curr_get_info.fifo_len =
 427                     config->tx_cfg[i].fifo_len - 1;
 428                 mac_control->fifos[i].fifo_no = i;
 429                 mac_control->fifos[i].nic = nic;
 430                 mac_control->fifos[i].max_txds = MAX_SKB_FRAGS + 2;
 431
 432                 for (j = 0; j < page_num; j++) {
 433                         int k = 0;
 434                         dma_addr_t tmp_p;
 435                         void *tmp_v;
 436                         tmp_v = pci_alloc_consistent(nic->pdev,
 437                                                      PAGE_SIZE, &tmp_p);
 438                         if (!tmp_v) {
 439                                 DBG_PRINT(ERR_DBG,
 440                                           "pci_alloc_consistent ");
 441                                 DBG_PRINT(ERR_DBG, "failed for TxDL\n");
 442                                 return -ENOMEM;
 443                         }
 444                         /* If we got a zero DMA address(can happen on
 445                          * certain platforms like PPC), reallocate.
 446                          * Store virtual address of page we don't want,
 447                          * to be freed later.
 448                          */
 449                         if (!tmp_p) {
 450                                 mac_control->zerodma_virt_addr = tmp_v;
 451                                 DBG_PRINT(INIT_DBG,
 452                                 "%s: Zero DMA address for TxDL. ", dev->name);
 453                                 DBG_PRINT(INIT_DBG,
 454                                 "Virtual address %p\n", tmp_v);
 455                                 tmp_v = pci_alloc_consistent(nic->pdev,
 456                                                      PAGE_SIZE, &tmp_p);
 457                                 if (!tmp_v) {
 458                                         DBG_PRINT(ERR_DBG,
 459                                           "pci_alloc_consistent ");
 460                                         DBG_PRINT(ERR_DBG, "failed for TxDL\n");
 461                                         return -ENOMEM;
 462                                 }
 463                         }
 464                         while (k < lst_per_page) {
 465                                 int l = (j * lst_per_page) + k;
 466                                 if (l == config->tx_cfg[i].fifo_len)
 467                                         break;
 468                                 mac_control->fifos[i].list_info[l].list_virt_addr =
 469                                     tmp_v + (k * lst_size);
 470                                 mac_control->fifos[i].list_info[l].list_phy_addr =
 471                                     tmp_p + (k * lst_size);
 472                                 k++;
 473                         }
 474                 }
 475         }
 476
 477         nic->ufo_in_band_v = kmalloc((sizeof(u64) * size), GFP_KERNEL);
 478         if (!nic->ufo_in_band_v)
 479                 return -ENOMEM;
 480
 481         /* Allocation and initialization of RXDs in Rings */
 482         size = 0;
 483         for (i = 0; i < config->rx_ring_num; i++) {
 484                 if (config->rx_cfg[i].num_rxd %
 485                     (rxd_count[nic->rxd_mode] + 1)) {
 486                         DBG_PRINT(ERR_DBG, "%s: RxD count of ", dev->name);
 487                         DBG_PRINT(ERR_DBG, "Ring%d is not a multiple of ",
 488                                   i);
 489                         DBG_PRINT(ERR_DBG, "RxDs per Block");
 490                         return FAILURE;
 491                 }
 492                 size += config->rx_cfg[i].num_rxd;
 493                 mac_control->rings[i].block_count =
 494                         config->rx_cfg[i].num_rxd /
 495                         (rxd_count[nic->rxd_mode] + 1 );
 496                 mac_control->rings[i].pkt_cnt = config->rx_cfg[i].num_rxd -
 497                         mac_control->rings[i].block_count;
 498         }
 499         if (nic->rxd_mode == RXD_MODE_1)
 500                 size = (size * (sizeof(RxD1_t)));
 501         else
 502                 size = (size * (sizeof(RxD3_t)));
 503         rx_sz = size;
 504
 505         for (i = 0; i < config->rx_ring_num; i++) {
 506                 mac_control->rings[i].rx_curr_get_info.block_index = 0;
 507                 mac_control->rings[i].rx_curr_get_info.offset = 0;
 508                 mac_control->rings[i].rx_curr_get_info.ring_len =
 509                     config->rx_cfg[i].num_rxd - 1;
 510                 mac_control->rings[i].rx_curr_put_info.block_index = 0;
 511                 mac_control->rings[i].rx_curr_put_info.offset = 0;
 512                 mac_control->rings[i].rx_curr_put_info.ring_len =
 513                     config->rx_cfg[i].num_rxd - 1;
 514                 mac_control->rings[i].nic = nic;
 515                 mac_control->rings[i].ring_no = i;
 516
 517                 blk_cnt = config->rx_cfg[i].num_rxd /
 518                                 (rxd_count[nic->rxd_mode] + 1);
 519                 /*  Allocating all the Rx blocks */
 520                 for (j = 0; j < blk_cnt; j++) {
 521                         rx_block_info_t *rx_blocks;
 522                         int l;
 523
 524                         rx_blocks = &mac_control->rings[i].rx_blocks[j];
 525                         size = SIZE_OF_BLOCK; //size is always page size
 526                         tmp_v_addr = pci_alloc_consistent(nic->pdev, size,
 527                                                           &tmp_p_addr);
 528                         if (tmp_v_addr == NULL) {
 529                                 /*
 530                                  * In case of failure, free_shared_mem()
 531                                  * is called, which should free any
 532                                  * memory that was alloced till the
 533                                  * failure happened.
 534                                  */
 535                                 rx_blocks->block_virt_addr = tmp_v_addr;
 536                                 return -ENOMEM;
 537                         }
 538                         memset(tmp_v_addr, 0, size);
 539                         rx_blocks->block_virt_addr = tmp_v_addr;
 540                         rx_blocks->block_dma_addr = tmp_p_addr;
 541                         rx_blocks->rxds = kmalloc(sizeof(rxd_info_t)*
 542                                                   rxd_count[nic->rxd_mode],
 543                                                   GFP_KERNEL);
 544                         for (l=0; l<rxd_count[nic->rxd_mode];l++) {
 545                                 rx_blocks->rxds[l].virt_addr =
 546                                         rx_blocks->block_virt_addr +
 547                                         (rxd_size[nic->rxd_mode] * l);
 548                                 rx_blocks->rxds[l].dma_addr =
 549                                         rx_blocks->block_dma_addr +
 550                                         (rxd_size[nic->rxd_mode] * l);
 551                         }
 552
 553                         mac_control->rings[i].rx_blocks[j].block_virt_addr =
 554                                 tmp_v_addr;
 555                         mac_control->rings[i].rx_blocks[j].block_dma_addr =
 556                                 tmp_p_addr;
 557                 }
 558                 /* Interlinking all Rx Blocks */
 559                 for (j = 0; j < blk_cnt; j++) {
 560                         tmp_v_addr =
 561                                 mac_control->rings[i].rx_blocks[j].block_virt_addr;
 562                         tmp_v_addr_next =
 563                                 mac_control->rings[i].rx_blocks[(j + 1) %
 564                                               blk_cnt].block_virt_addr;
 565                         tmp_p_addr =
 566                                 mac_control->rings[i].rx_blocks[j].block_dma_addr;
 567                         tmp_p_addr_next =
 568                                 mac_control->rings[i].rx_blocks[(j + 1) %
 569                                               blk_cnt].block_dma_addr;
 570
 571                         pre_rxd_blk = (RxD_block_t *) tmp_v_addr;
 572                         pre_rxd_blk->reserved_2_pNext_RxD_block =
 573                             (unsigned long) tmp_v_addr_next;
 574                         pre_rxd_blk->pNext_RxD_Blk_physical =
 575                             (u64) tmp_p_addr_next;
 576                 }
 577         }
 578         if (nic->rxd_mode >= RXD_MODE_3A) {
 579                 /*
 580                  * Allocation of Storages for buffer addresses in 2BUFF mode
 581                  * and the buffers as well.
 582                  */
 583                 for (i = 0; i < config->rx_ring_num; i++) {
 584                         blk_cnt = config->rx_cfg[i].num_rxd /
 585                            (rxd_count[nic->rxd_mode]+ 1);
 586                         mac_control->rings[i].ba =
 587                                 kmalloc((sizeof(buffAdd_t *) * blk_cnt),
 588                                      GFP_KERNEL);
 589                         if (!mac_control->rings[i].ba)
 590                                 return -ENOMEM;
 591                         for (j = 0; j < blk_cnt; j++) {
 592                                 int k = 0;
 593                                 mac_control->rings[i].ba[j] =
 594                                         kmalloc((sizeof(buffAdd_t) *
 595                                                 (rxd_count[nic->rxd_mode] + 1)),
 596                                                 GFP_KERNEL);
 597                                 if (!mac_control->rings[i].ba[j])
 598                                         return -ENOMEM;
 599                                 while (k != rxd_count[nic->rxd_mode]) {
 600                                         ba = &mac_control->rings[i].ba[j][k];
 601
 602                                         ba->ba_0_org = (void *) kmalloc
 603                                             (BUF0_LEN + ALIGN_SIZE, GFP_KERNEL);
 604                                         if (!ba->ba_0_org)
 605                                                 return -ENOMEM;
 606                                         tmp = (unsigned long)ba->ba_0_org;
 607                                         tmp += ALIGN_SIZE;
 608                                         tmp &= ~((unsigned long) ALIGN_SIZE);
 609                                         ba->ba_0 = (void *) tmp;
 610
 611                                         ba->ba_1_org = (void *) kmalloc
 612                                             (BUF1_LEN + ALIGN_SIZE, GFP_KERNEL);
 613                                         if (!ba->ba_1_org)
 614                                                 return -ENOMEM;
 615                                         tmp = (unsigned long) ba->ba_1_org;
 616                                         tmp += ALIGN_SIZE;
 617                                         tmp &= ~((unsigned long) ALIGN_SIZE);
 618                                         ba->ba_1 = (void *) tmp;
 619                                         k++;
 620                                 }
 621                         }
 622                 }
 623         }
 624
 625         /* Allocation and initialization of Statistics block */
 626         size = sizeof(StatInfo_t);
 627         mac_control->stats_mem = pci_alloc_consistent
 628             (nic->pdev, size, &mac_control->stats_mem_phy);
 629
 630         if (!mac_control->stats_mem) {
 631                 /*
 632                  * In case of failure, free_shared_mem() is called, which
 633                  * should free any memory that was alloced till the
 634                  * failure happened.
 635                  */
 636                 return -ENOMEM;
 637         }
 638         mac_control->stats_mem_sz = size;
 639
 640         tmp_v_addr = mac_control->stats_mem;
 641         mac_control->stats_info = (StatInfo_t *) tmp_v_addr;
 642         memset(tmp_v_addr, 0, size);
 643         DBG_PRINT(INIT_DBG, "%s:Ring Mem PHY: 0x%llx\n", dev->name,
 644                   (unsigned long long) tmp_p_addr);
 645
 646         return SUCCESS;
 647 }
 648
 649 /**
 650  * free_shared_mem - Free the allocated Memory
 651  * @nic:  Device private variable.
 652  * Description: This function is to free all memory locations allocated by
 653  * the init_shared_mem() function and return it to the kernel.
 654  */
 655
 656 static void free_shared_mem(struct s2io_nic *nic)
 657 {
 658         int i, j, blk_cnt, size;
 659         void *tmp_v_addr;
 660         dma_addr_t tmp_p_addr;
 661         mac_info_t *mac_control;
 662         struct config_param *config;
 663         int lst_size, lst_per_page;
 664         struct net_device *dev = nic->dev;
 665
 666         if (!nic)
 667                 return;
 668
 669         mac_control = &nic->mac_control;
 670         config = &nic->config;
 671
 672         lst_size = (sizeof(TxD_t) * config->max_txds);
 673         lst_per_page = PAGE_SIZE / lst_size;
 674
 675         for (i = 0; i < config->tx_fifo_num; i++) {
 676                 int page_num = TXD_MEM_PAGE_CNT(config->tx_cfg[i].fifo_len,
 677                                                 lst_per_page);
 678                 for (j = 0; j < page_num; j++) {
 679                         int mem_blks = (j * lst_per_page);
 680                         if (!mac_control->fifos[i].list_info)
 681                                 return;
 682                         if (!mac_control->fifos[i].list_info[mem_blks].
 683                                  list_virt_addr)
 684                                 break;
 685                         pci_free_consistent(nic->pdev, PAGE_SIZE,
 686                                             mac_control->fifos[i].
 687                                             list_info[mem_blks].
 688                                             list_virt_addr,
 689                                             mac_control->fifos[i].
 690                                             list_info[mem_blks].
 691                                             list_phy_addr);
 692                 }
 693                 /* If we got a zero DMA address during allocation,
 694                  * free the page now
 695                  */
 696                 if (mac_control->zerodma_virt_addr) {
 697                         pci_free_consistent(nic->pdev, PAGE_SIZE,
 698                                             mac_control->zerodma_virt_addr,
 699                                             (dma_addr_t)0);
 700                         DBG_PRINT(INIT_DBG,
 701                                 "%s: Freeing TxDL with zero DMA addr. ",
 702                                 dev->name);
 703                         DBG_PRINT(INIT_DBG, "Virtual address %p\n",
 704                                 mac_control->zerodma_virt_addr);
 705                 }
 706                 kfree(mac_control->fifos[i].list_info);
 707         }
 708
 709         size = SIZE_OF_BLOCK;
 710         for (i = 0; i < config->rx_ring_num; i++) {
 711                 blk_cnt = mac_control->rings[i].block_count;
 712                 for (j = 0; j < blk_cnt; j++) {
 713                         tmp_v_addr = mac_control->rings[i].rx_blocks[j].
 714                                 block_virt_addr;
 715                         tmp_p_addr = mac_control->rings[i].rx_blocks[j].
 716                                 block_dma_addr;
 717                         if (tmp_v_addr == NULL)
 718                                 break;
 719                         pci_free_consistent(nic->pdev, size,
 720                                             tmp_v_addr, tmp_p_addr);
 721                         kfree(mac_control->rings[i].rx_blocks[j].rxds);
 722                 }
 723         }
 724
 725         if (nic->rxd_mode >= RXD_MODE_3A) {
 726                 /* Freeing buffer storage addresses in 2BUFF mode. */
 727                 for (i = 0; i < config->rx_ring_num; i++) {
 728                         blk_cnt = config->rx_cfg[i].num_rxd /
 729                             (rxd_count[nic->rxd_mode] + 1);
 730                         for (j = 0; j < blk_cnt; j++) {
 731                                 int k = 0;
 732                                 if (!mac_control->rings[i].ba[j])
 733                                         continue;
 734                                 while (k != rxd_count[nic->rxd_mode]) {
 735                                         buffAdd_t *ba =
 736                                                 &mac_control->rings[i].ba[j][k];
 737                                         kfree(ba->ba_0_org);
 738                                         kfree(ba->ba_1_org);
 739                                         k++;
 740                                 }
 741                                 kfree(mac_control->rings[i].ba[j]);
 742                         }
 743                         kfree(mac_control->rings[i].ba);
 744                 }
 745         }
 746
 747         if (mac_control->stats_mem) {
 748                 pci_free_consistent(nic->pdev,
 749                                     mac_control->stats_mem_sz,
 750                                     mac_control->stats_mem,
 751                                     mac_control->stats_mem_phy);
 752         }
 753         if (nic->ufo_in_band_v)
 754                 kfree(nic->ufo_in_band_v);
 755 }
 756
 757 /**
 758  * s2io_verify_pci_mode -
 759  */
 760
 761 static int s2io_verify_pci_mode(nic_t *nic)
 762 {
 763         XENA_dev_config_t __iomem *bar0 = nic->bar0;
 764         register u64 val64 = 0;
 765         int     mode;
 766
 767         val64 = readq(&bar0->pci_mode);
 768         mode = (u8)GET_PCI_MODE(val64);
 769
 770         if ( val64 & PCI_MODE_UNKNOWN_MODE)
 771                 return -1;      /* Unknown PCI mode */
 772         return mode;
 773 }
 774
 775
 776 /**
 777  * s2io_print_pci_mode -
 778  */
 779 static int s2io_print_pci_mode(nic_t *nic)
 780 {
 781         XENA_dev_config_t __iomem *bar0 = nic->bar0;
 782         register u64 val64 = 0;
 783         int     mode;
 784         struct config_param *config = &nic->config;
 785
 786         val64 = readq(&bar0->pci_mode);
 787         mode = (u8)GET_PCI_MODE(val64);
 788
 789         if ( val64 & PCI_MODE_UNKNOWN_MODE)
 790                 return -1;      /* Unknown PCI mode */
 791
 792         if (val64 & PCI_MODE_32_BITS) {
 793                 DBG_PRINT(ERR_DBG, "%s: Device is on 32 bit ", nic->dev->name);
 794         } else {
 795                 DBG_PRINT(ERR_DBG, "%s: Device is on 64 bit ", nic->dev->name);
 796         }
 797
 798         switch(mode) {
 799                 case PCI_MODE_PCI_33:
 800                         DBG_PRINT(ERR_DBG, "33MHz PCI bus\n");
 801                         config->bus_speed = 33;
 802                         break;
 803                 case PCI_MODE_PCI_66:
 804                         DBG_PRINT(ERR_DBG, "66MHz PCI bus\n");
 805                         config->bus_speed = 133;
 806                         break;
 807                 case PCI_MODE_PCIX_M1_66:
 808                         DBG_PRINT(ERR_DBG, "66MHz PCIX(M1) bus\n");
 809                         config->bus_speed = 133; /* Herc doubles the clock rate */
 810                         break;
 811                 case PCI_MODE_PCIX_M1_100:
 812                         DBG_PRINT(ERR_DBG, "100MHz PCIX(M1) bus\n");
 813                         config->bus_speed = 200;
 814                         break;
 815                 case PCI_MODE_PCIX_M1_133:
 816                         DBG_PRINT(ERR_DBG, "133MHz PCIX(M1) bus\n");
 817                         config->bus_speed = 266;
 818                         break;
 819                 case PCI_MODE_PCIX_M2_66:
 820                         DBG_PRINT(ERR_DBG, "133MHz PCIX(M2) bus\n");
 821                         config->bus_speed = 133;
 822                         break;
 823                 case PCI_MODE_PCIX_M2_100:
 824                         DBG_PRINT(ERR_DBG, "200MHz PCIX(M2) bus\n");
 825                         config->bus_speed = 200;
 826                         break;
 827                 case PCI_MODE_PCIX_M2_133:
 828                         DBG_PRINT(ERR_DBG, "266MHz PCIX(M2) bus\n");
 829                         config->bus_speed = 266;
 830                         break;
 831                 default:
 832                         return -1;      /* Unsupported bus speed */
 833         }
 834
 835         return mode;
 836 }
 837
 838 /**
 839  *  init_nic - Initialization of hardware
 840  *  @nic: device peivate variable
 841  *  Description: The function sequentially configures every block
 842  *  of the H/W from their reset values.
 843  *  Return Value:  SUCCESS on success and
 844  *  '-1' on failure (endian settings incorrect).
 845  */
 846
 847 static int init_nic(struct s2io_nic *nic)
 848 {
 849         XENA_dev_config_t __iomem *bar0 = nic->bar0;
 850         struct net_device *dev = nic->dev;
 851         register u64 val64 = 0;
 852         void __iomem *add;
 853         u32 time;
 854         int i, j;
 855         mac_info_t *mac_control;
 856         struct config_param *config;
 857         int mdio_cnt = 0, dtx_cnt = 0;
 858         unsigned long long mem_share;
 859         int mem_size;
 860
 861         mac_control = &nic->mac_control;
 862         config = &nic->config;
 863
 864         /* to set the swapper controle on the card */
 865         if(s2io_set_swapper(nic)) {
 866                 DBG_PRINT(ERR_DBG,"ERROR: Setting Swapper failed\n");
 867                 return -1;
 868         }
 869
 870         /*
 871          * Herc requires EOI to be removed from reset before XGXS, so..
 872          */
 873         if (nic->device_type & XFRAME_II_DEVICE) {
 874                 val64 = 0xA500000000ULL;
 875                 writeq(val64, &bar0->sw_reset);
 876                 msleep(500);
 877                 val64 = readq(&bar0->sw_reset);
 878         }
 879
 880         /* Remove XGXS from reset state */
 881         val64 = 0;
 882         writeq(val64, &bar0->sw_reset);
 883         msleep(500);
 884         val64 = readq(&bar0->sw_reset);
 885
 886         /*  Enable Receiving broadcasts */
 887         add = &bar0->mac_cfg;
 888         val64 = readq(&bar0->mac_cfg);
 889         val64 |= MAC_RMAC_BCAST_ENABLE;
 890         writeq(RMAC_CFG_KEY(0x4C0D), &bar0->rmac_cfg_key);
 891         writel((u32) val64, add);
 892         writeq(RMAC_CFG_KEY(0x4C0D), &bar0->rmac_cfg_key);
 893         writel((u32) (val64 >> 32), (add + 4));
 894
 895         /* Read registers in all blocks */
 896         val64 = readq(&bar0->mac_int_mask);
 897         val64 = readq(&bar0->mc_int_mask);
 898         val64 = readq(&bar0->xgxs_int_mask);
 899
 900         /*  Set MTU */
 901         val64 = dev->mtu;
 902         writeq(vBIT(val64, 2, 14), &bar0->rmac_max_pyld_len);
 903
 904         /*
 905          * Configuring the XAUI Interface of Xena.
 906          * ***************************************
 907          * To Configure the Xena's XAUI, one has to write a series
 908          * of 64 bit values into two registers in a particular
 909          * sequence. Hence a macro 'SWITCH_SIGN' has been defined
 910          * which will be defined in the array of configuration values
 911          * (xena_dtx_cfg & xena_mdio_cfg) at appropriate places
 912          * to switch writing from one regsiter to another. We continue
 913          * writing these values until we encounter the 'END_SIGN' macro.
 914          * For example, After making a series of 21 writes into
 915          * dtx_control register the 'SWITCH_SIGN' appears and hence we
 916          * start writing into mdio_control until we encounter END_SIGN.
 917          */
 918         if (nic->device_type & XFRAME_II_DEVICE) {
 919                 while (herc_act_dtx_cfg[dtx_cnt] != END_SIGN) {
 920                         SPECIAL_REG_WRITE(herc_act_dtx_cfg[dtx_cnt],
 921                                           &bar0->dtx_control, UF);
 922                         if (dtx_cnt & 0x1)
 923                                 msleep(1); /* Necessary!! */
 924                         dtx_cnt++;
 925                 }
 926         } else {
 927                 while (1) {
 928                       dtx_cfg:
 929                         while (xena_dtx_cfg[dtx_cnt] != END_SIGN) {
 930                                 if (xena_dtx_cfg[dtx_cnt] == SWITCH_SIGN) {
 931                                         dtx_cnt++;
 932                                         goto mdio_cfg;
 933                                 }
 934                                 SPECIAL_REG_WRITE(xena_dtx_cfg[dtx_cnt],
 935                                                   &bar0->dtx_control, UF);
 936                                 val64 = readq(&bar0->dtx_control);
 937                                 dtx_cnt++;
 938                         }
 939                       mdio_cfg:
 940                         while (xena_mdio_cfg[mdio_cnt] != END_SIGN) {
 941                                 if (xena_mdio_cfg[mdio_cnt] == SWITCH_SIGN) {
 942                                         mdio_cnt++;
 943                                         goto dtx_cfg;
 944                                 }
 945                                 SPECIAL_REG_WRITE(xena_mdio_cfg[mdio_cnt],
 946                                                   &bar0->mdio_control, UF);
 947                                 val64 = readq(&bar0->mdio_control);
 948                                 mdio_cnt++;
 949                         }
 950                         if ((xena_dtx_cfg[dtx_cnt] == END_SIGN) &&
 951                             (xena_mdio_cfg[mdio_cnt] == END_SIGN)) {
 952                                 break;
 953                         } else {
 954                                 goto dtx_cfg;
 955                         }
 956                 }
 957         }
 958
 959         /*  Tx DMA Initialization */
 960         val64 = 0;
 961         writeq(val64, &bar0->tx_fifo_partition_0);
 962         writeq(val64, &bar0->tx_fifo_partition_1);
 963         writeq(val64, &bar0->tx_fifo_partition_2);
 964         writeq(val64, &bar0->tx_fifo_partition_3);
 965
 966
 967         for (i = 0, j = 0; i < config->tx_fifo_num; i++) {
 968                 val64 |=
 969                     vBIT(config->tx_cfg[i].fifo_len - 1, ((i * 32) + 19),
 970                          13) | vBIT(config->tx_cfg[i].fifo_priority,
 971                                     ((i * 32) + 5), 3);
 972
 973                 if (i == (config->tx_fifo_num - 1)) {
 974                         if (i % 2 == 0)
 975                                 i++;
 976                 }
 977
 978                 switch (i) {
 979                 case 1:
 980                         writeq(val64, &bar0->tx_fifo_partition_0);
 981                         val64 = 0;
 982                         break;
 983                 case 3:
 984                         writeq(val64, &bar0->tx_fifo_partition_1);
 985                         val64 = 0;
 986                         break;
 987                 case 5:
 988                         writeq(val64, &bar0->tx_fifo_partition_2);
 989                         val64 = 0;
 990                         break;
 991                 case 7:
 992                         writeq(val64, &bar0->tx_fifo_partition_3);
 993                         break;
 994                 }
 995         }
 996
 997         /* Enable Tx FIFO partition 0. */
 998         val64 = readq(&bar0->tx_fifo_partition_0);
 999         val64 |= BIT(0);        /* To enable the FIFO partition. */
1000         writeq(val64, &bar0->tx_fifo_partition_0);
1001
1002         /*
1003          * Disable 4 PCCs for Xena1, 2 and 3 as per H/W bug
1004          * SXE-008 TRANSMIT DMA ARBITRATION ISSUE.
1005          */
1006         if ((nic->device_type == XFRAME_I_DEVICE) &&
1007                 (get_xena_rev_id(nic->pdev) < 4))
1008                 writeq(PCC_ENABLE_FOUR, &bar0->pcc_enable);
1009
1010         val64 = readq(&bar0->tx_fifo_partition_0);
1011         DBG_PRINT(INIT_DBG, "Fifo partition at: 0x%p is: 0x%llx\n",
1012                   &bar0->tx_fifo_partition_0, (unsigned long long) val64);
1013
1014         /*
1015          * Initialization of Tx_PA_CONFIG register to ignore packet
1016          * integrity checking.
1017          */
1018         val64 = readq(&bar0->tx_pa_cfg);
1019         val64 |= TX_PA_CFG_IGNORE_FRM_ERR | TX_PA_CFG_IGNORE_SNAP_OUI |
1020             TX_PA_CFG_IGNORE_LLC_CTRL | TX_PA_CFG_IGNORE_L2_ERR;
1021         writeq(val64, &bar0->tx_pa_cfg);
1022
1023         /* Rx DMA intialization. */
1024         val64 = 0;
1025         for (i = 0; i < config->rx_ring_num; i++) {
1026                 val64 |=
1027                     vBIT(config->rx_cfg[i].ring_priority, (5 + (i * 8)),
1028                          3);
1029         }
1030         writeq(val64, &bar0->rx_queue_priority);
1031
1032         /*
1033          * Allocating equal share of memory to all the
1034          * configured Rings.
1035          */
1036         val64 = 0;
1037         if (nic->device_type & XFRAME_II_DEVICE)
1038                 mem_size = 32;
1039         else
1040                 mem_size = 64;
1041
1042         for (i = 0; i < config->rx_ring_num; i++) {
1043                 switch (i) {
1044                 case 0:
1045                         mem_share = (mem_size / config->rx_ring_num +
1046                                      mem_size % config->rx_ring_num);
1047                         val64 |= RX_QUEUE_CFG_Q0_SZ(mem_share);
1048                         continue;
1049                 case 1:
1050                         mem_share = (mem_size / config->rx_ring_num);
1051                         val64 |= RX_QUEUE_CFG_Q1_SZ(mem_share);
1052                         continue;
1053                 case 2:
1054                         mem_share = (mem_size / config->rx_ring_num);
1055                         val64 |= RX_QUEUE_CFG_Q2_SZ(mem_share);
1056                         continue;
1057                 case 3:
1058                         mem_share = (mem_size / config->rx_ring_num);
1059                         val64 |= RX_QUEUE_CFG_Q3_SZ(mem_share);
1060                         continue;
1061                 case 4:
1062                         mem_share = (mem_size / config->rx_ring_num);
1063                         val64 |= RX_QUEUE_CFG_Q4_SZ(mem_share);
1064                         continue;
1065                 case 5:
1066                         mem_share = (mem_size / config->rx_ring_num);
1067                         val64 |= RX_QUEUE_CFG_Q5_SZ(mem_share);
1068                         continue;
1069                 case 6:
1070                         mem_share = (mem_size / config->rx_ring_num);
1071                         val64 |= RX_QUEUE_CFG_Q6_SZ(mem_share);
1072                         continue;
1073                 case 7:
1074                         mem_share = (mem_size / config->rx_ring_num);
1075                         val64 |= RX_QUEUE_CFG_Q7_SZ(mem_share);
1076                         continue;
1077                 }
1078         }
1079         writeq(val64, &bar0->rx_queue_cfg);
1080
1081         /*
1082          * Filling Tx round robin registers
1083          * as per the number of FIFOs
1084          */
1085         switch (config->tx_fifo_num) {
1086         case 1:
1087                 val64 = 0x0000000000000000ULL;
1088                 writeq(val64, &bar0->tx_w_round_robin_0);
1089                 writeq(val64, &bar0->tx_w_round_robin_1);
1090                 writeq(val64, &bar0->tx_w_round_robin_2);
1091                 writeq(val64, &bar0->tx_w_round_robin_3);
1092                 writeq(val64, &bar0->tx_w_round_robin_4);
1093                 break;
1094         case 2:
1095                 val64 = 0x0000010000010000ULL;
1096                 writeq(val64, &bar0->tx_w_round_robin_0);
1097                 val64 = 0x0100000100000100ULL;
1098                 writeq(val64, &bar0->tx_w_round_robin_1);
1099                 val64 = 0x0001000001000001ULL;
1100                 writeq(val64, &bar0->tx_w_round_robin_2);
1101                 val64 = 0x0000010000010000ULL;
1102                 writeq(val64, &bar0->tx_w_round_robin_3);
1103                 val64 = 0x0100000000000000ULL;
1104                 writeq(val64, &bar0->tx_w_round_robin_4);
1105                 break;
1106         case 3:
1107                 val64 = 0x0001000102000001ULL;
1108                 writeq(val64, &bar0->tx_w_round_robin_0);
1109                 val64 = 0x0001020000010001ULL;
1110                 writeq(val64, &bar0->tx_w_round_robin_1);
1111                 val64 = 0x0200000100010200ULL;
1112                 writeq(val64, &bar0->tx_w_round_robin_2);
1113                 val64 = 0x0001000102000001ULL;
1114                 writeq(val64, &bar0->tx_w_round_robin_3);
1115                 val64 = 0x0001020000000000ULL;
1116                 writeq(val64, &bar0->tx_w_round_robin_4);
1117                 break;
1118         case 4:
1119                 val64 = 0x0001020300010200ULL;
1120                 writeq(val64, &bar0->tx_w_round_robin_0);
1121                 val64 = 0x0100000102030001ULL;
1122                 writeq(val64, &bar0->tx_w_round_robin_1);
1123                 val64 = 0x0200010000010203ULL;
1124                 writeq(val64, &bar0->tx_w_round_robin_2);
1125                 val64 = 0x0001020001000001ULL;
1126                 writeq(val64, &bar0->tx_w_round_robin_3);
1127                 val64 = 0x0203000100000000ULL;
1128                 writeq(val64, &bar0->tx_w_round_robin_4);
1129                 break;
1130         case 5:
1131                 val64 = 0x0001000203000102ULL;
1132                 writeq(val64, &bar0->tx_w_round_robin_0);
1133                 val64 = 0x0001020001030004ULL;
1134                 writeq(val64, &bar0->tx_w_round_robin_1);
1135                 val64 = 0x0001000203000102ULL;
1136                 writeq(val64, &bar0->tx_w_round_robin_2);
1137                 val64 = 0x0001020001030004ULL;
1138                 writeq(val64, &bar0->tx_w_round_robin_3);
1139                 val64 = 0x0001000000000000ULL;
1140                 writeq(val64, &bar0->tx_w_round_robin_4);
1141                 break;
1142         case 6:
1143                 val64 = 0x0001020304000102ULL;
1144                 writeq(val64, &bar0->tx_w_round_robin_0);
1145                 val64 = 0x0304050001020001ULL;
1146                 writeq(val64, &bar0->tx_w_round_robin_1);
1147                 val64 = 0x0203000100000102ULL;
1148                 writeq(val64, &bar0->tx_w_round_robin_2);
1149                 val64 = 0x0304000102030405ULL;
1150                 writeq(val64, &bar0->tx_w_round_robin_3);
1151                 val64 = 0x0001000200000000ULL;
1152                 writeq(val64, &bar0->tx_w_round_robin_4);
1153                 break;
1154         case 7:
1155                 val64 = 0x0001020001020300ULL;
1156                 writeq(val64, &bar0->tx_w_round_robin_0);
1157                 val64 = 0x0102030400010203ULL;
1158                 writeq(val64, &bar0->tx_w_round_robin_1);
1159                 val64 = 0x0405060001020001ULL;
1160                 writeq(val64, &bar0->tx_w_round_robin_2);
1161                 val64 = 0x0304050000010200ULL;
1162                 writeq(val64, &bar0->tx_w_round_robin_3);
1163                 val64 = 0x0102030000000000ULL;
1164                 writeq(val64, &bar0->tx_w_round_robin_4);
1165                 break;
1166         case 8:
1167                 val64 = 0x0001020300040105ULL;
1168                 writeq(val64, &bar0->tx_w_round_robin_0);
1169                 val64 = 0x0200030106000204ULL;
1170                 writeq(val64, &bar0->tx_w_round_robin_1);
1171                 val64 = 0x0103000502010007ULL;
1172                 writeq(val64, &bar0->tx_w_round_robin_2);
1173                 val64 = 0x0304010002060500ULL;
1174                 writeq(val64, &bar0->tx_w_round_robin_3);
1175                 val64 = 0x0103020400000000ULL;
1176                 writeq(val64, &bar0->tx_w_round_robin_4);
1177                 break;
1178         }
1179
1180         /* Filling the Rx round robin registers as per the
1181          * number of Rings and steering based on QoS.
1182          */
1183         switch (config->rx_ring_num) {
1184         case 1:
1185                 val64 = 0x8080808080808080ULL;
1186                 writeq(val64, &bar0->rts_qos_steering);
1187                 break;
1188         case 2:
1189                 val64 = 0x0000010000010000ULL;
1190                 writeq(val64, &bar0->rx_w_round_robin_0);
1191                 val64 = 0x0100000100000100ULL;
1192                 writeq(val64, &bar0->rx_w_round_robin_1);
1193                 val64 = 0x0001000001000001ULL;
1194                 writeq(val64, &bar0->rx_w_round_robin_2);
1195                 val64 = 0x0000010000010000ULL;
1196                 writeq(val64, &bar0->rx_w_round_robin_3);
1197                 val64 = 0x0100000000000000ULL;
1198                 writeq(val64, &bar0->rx_w_round_robin_4);
1199
1200                 val64 = 0x8080808040404040ULL;
1201                 writeq(val64, &bar0->rts_qos_steering);
1202                 break;
1203         case 3:
1204                 val64 = 0x0001000102000001ULL;
1205                 writeq(val64, &bar0->rx_w_round_robin_0);
1206                 val64 = 0x0001020000010001ULL;
1207                 writeq(val64, &bar0->rx_w_round_robin_1);
1208                 val64 = 0x0200000100010200ULL;
1209                 writeq(val64, &bar0->rx_w_round_robin_2);
1210                 val64 = 0x0001000102000001ULL;
1211                 writeq(val64, &bar0->rx_w_round_robin_3);
1212                 val64 = 0x0001020000000000ULL;
1213                 writeq(val64, &bar0->rx_w_round_robin_4);
1214
1215                 val64 = 0x8080804040402020ULL;
1216                 writeq(val64, &bar0->rts_qos_steering);
1217                 break;
1218         case 4:
1219                 val64 = 0x0001020300010200ULL;
1220                 writeq(val64, &bar0->rx_w_round_robin_0);
1221                 val64 = 0x0100000102030001ULL;
1222                 writeq(val64, &bar0->rx_w_round_robin_1);
1223                 val64 = 0x0200010000010203ULL;
1224                 writeq(val64, &bar0->rx_w_round_robin_2);
1225                 val64 = 0x0001020001000001ULL;
1226                 writeq(val64, &bar0->rx_w_round_robin_3);
1227                 val64 = 0x0203000100000000ULL;
1228                 writeq(val64, &bar0->rx_w_round_robin_4);
1229
1230                 val64 = 0x8080404020201010ULL;
1231                 writeq(val64, &bar0->rts_qos_steering);
1232                 break;
1233         case 5:
1234                 val64 = 0x0001000203000102ULL;
1235                 writeq(val64, &bar0->rx_w_round_robin_0);
1236                 val64 = 0x0001020001030004ULL;
1237                 writeq(val64, &bar0->rx_w_round_robin_1);
1238                 val64 = 0x0001000203000102ULL;
1239                 writeq(val64, &bar0->rx_w_round_robin_2);
1240                 val64 = 0x0001020001030004ULL;
1241                 writeq(val64, &bar0->rx_w_round_robin_3);
1242                 val64 = 0x0001000000000000ULL;
1243                 writeq(val64, &bar0->rx_w_round_robin_4);
1244
1245                 val64 = 0x8080404020201008ULL;
1246                 writeq(val64, &bar0->rts_qos_steering);
1247                 break;
1248         case 6:
1249                 val64 = 0x0001020304000102ULL;
1250                 writeq(val64, &bar0->rx_w_round_robin_0);
1251                 val64 = 0x0304050001020001ULL;
1252                 writeq(val64, &bar0->rx_w_round_robin_1);
1253                 val64 = 0x0203000100000102ULL;
1254                 writeq(val64, &bar0->rx_w_round_robin_2);
1255                 val64 = 0x0304000102030405ULL;
1256                 writeq(val64, &bar0->rx_w_round_robin_3);
1257                 val64 = 0x0001000200000000ULL;
1258                 writeq(val64, &bar0->rx_w_round_robin_4);
1259
1260                 val64 = 0x8080404020100804ULL;
1261                 writeq(val64, &bar0->rts_qos_steering);
1262                 break;
1263         case 7:
1264                 val64 = 0x0001020001020300ULL;
1265                 writeq(val64, &bar0->rx_w_round_robin_0);
1266                 val64 = 0x0102030400010203ULL;
1267                 writeq(val64, &bar0->rx_w_round_robin_1);
1268                 val64 = 0x0405060001020001ULL;
1269                 writeq(val64, &bar0->rx_w_round_robin_2);
1270                 val64 = 0x0304050000010200ULL;
1271                 writeq(val64, &bar0->rx_w_round_robin_3);
1272                 val64 = 0x0102030000000000ULL;
1273                 writeq(val64, &bar0->rx_w_round_robin_4);
1274
1275                 val64 = 0x8080402010080402ULL;
1276                 writeq(val64, &bar0->rts_qos_steering);
1277                 break;
1278         case 8:
1279                 val64 = 0x0001020300040105ULL;
1280                 writeq(val64, &bar0->rx_w_round_robin_0);
1281                 val64 = 0x0200030106000204ULL;
1282                 writeq(val64, &bar0->rx_w_round_robin_1);
1283                 val64 = 0x0103000502010007ULL;
1284                 writeq(val64, &bar0->rx_w_round_robin_2);
1285                 val64 = 0x0304010002060500ULL;
1286                 writeq(val64, &bar0->rx_w_round_robin_3);
1287                 val64 = 0x0103020400000000ULL;
1288                 writeq(val64, &bar0->rx_w_round_robin_4);
1289
1290                 val64 = 0x8040201008040201ULL;
1291                 writeq(val64, &bar0->rts_qos_steering);
1292                 break;
1293         }
1294
1295         /* UDP Fix */
1296         val64 = 0;
1297         for (i = 0; i < 8; i++)
1298                 writeq(val64, &bar0->rts_frm_len_n[i]);
1299
1300         /* Set the default rts frame length for the rings configured */
1301         val64 = MAC_RTS_FRM_LEN_SET(dev->mtu+22);
1302         for (i = 0 ; i < config->rx_ring_num ; i++)
1303                 writeq(val64, &bar0->rts_frm_len_n[i]);
1304
1305         /* Set the frame length for the configured rings
1306          * desired by the user
1307          */
1308         for (i = 0; i < config->rx_ring_num; i++) {
1309                 /* If rts_frm_len[i] == 0 then it is assumed that user not
1310                  * specified frame length steering.
1311                  * If the user provides the frame length then program
1312                  * the rts_frm_len register for those values or else
1313                  * leave it as it is.
1314                  */
1315                 if (rts_frm_len[i] != 0) {
1316                         writeq(MAC_RTS_FRM_LEN_SET(rts_frm_len[i]),
1317                                 &bar0->rts_frm_len_n[i]);
1318                 }
1319         }
1320
1321         /* Program statistics memory */
1322         writeq(mac_control->stats_mem_phy, &bar0->stat_addr);
1323
1324         if (nic->device_type == XFRAME_II_DEVICE) {
1325                 val64 = STAT_BC(0x320);
1326                 writeq(val64, &bar0->stat_byte_cnt);
1327         }
1328
1329         /*
1330          * Initializing the sampling rate for the device to calculate the
1331          * bandwidth utilization.
1332          */
1333         val64 = MAC_TX_LINK_UTIL_VAL(tmac_util_period) |
1334             MAC_RX_LINK_UTIL_VAL(rmac_util_period);
1335         writeq(val64, &bar0->mac_link_util);
1336
1337
1338         /*
1339          * Initializing the Transmit and Receive Traffic Interrupt
1340          * Scheme.
1341          */
1342         /*
1343          * TTI Initialization. Default Tx timer gets us about
1344          * 250 interrupts per sec. Continuous interrupts are enabled
1345          * by default.
1346          */
1347         if (nic->device_type == XFRAME_II_DEVICE) {
1348                 int count = (nic->config.bus_speed * 125)/2;
1349                 val64 = TTI_DATA1_MEM_TX_TIMER_VAL(count);
1350         } else {
1351
1352                 val64 = TTI_DATA1_MEM_TX_TIMER_VAL(0x2078);
1353         }
1354         val64 |= TTI_DATA1_MEM_TX_URNG_A(0xA) |
1355             TTI_DATA1_MEM_TX_URNG_B(0x10) |
1356             TTI_DATA1_MEM_TX_URNG_C(0x30) | TTI_DATA1_MEM_TX_TIMER_AC_EN;
1357                 if (use_continuous_tx_intrs)
1358                         val64 |= TTI_DATA1_MEM_TX_TIMER_CI_EN;
1359         writeq(val64, &bar0->tti_data1_mem);
1360
1361         val64 = TTI_DATA2_MEM_TX_UFC_A(0x10) |
1362             TTI_DATA2_MEM_TX_UFC_B(0x20) |
1363             TTI_DATA2_MEM_TX_UFC_C(0x70) | TTI_DATA2_MEM_TX_UFC_D(0x80);
1364         writeq(val64, &bar0->tti_data2_mem);
1365
1366         val64 = TTI_CMD_MEM_WE | TTI_CMD_MEM_STROBE_NEW_CMD;
1367         writeq(val64, &bar0->tti_command_mem);
1368
1369         /*
1370          * Once the operation completes, the Strobe bit of the command
1371          * register will be reset. We poll for this particular condition
1372          * We wait for a maximum of 500ms for the operation to complete,
1373          * if it's not complete by then we return error.
1374          */
1375         time = 0;
1376         while (TRUE) {
1377                 val64 = readq(&bar0->tti_command_mem);
1378                 if (!(val64 & TTI_CMD_MEM_STROBE_NEW_CMD)) {
1379                         break;
1380                 }
1381                 if (time > 10) {
1382                         DBG_PRINT(ERR_DBG, "%s: TTI init Failed\n",
1383                                   dev->name);
1384                         return -1;
1385                 }
1386                 msleep(50);
1387                 time++;
1388         }
1389
1390         if (nic->config.bimodal) {
1391                 int k = 0;
1392                 for (k = 0; k < config->rx_ring_num; k++) {
1393                         val64 = TTI_CMD_MEM_WE | TTI_CMD_MEM_STROBE_NEW_CMD;
1394                         val64 |= TTI_CMD_MEM_OFFSET(0x38+k);
1395                         writeq(val64, &bar0->tti_command_mem);
1396
1397                 /*
1398                  * Once the operation completes, the Strobe bit of the command
1399                  * register will be reset. We poll for this particular condition
1400                  * We wait for a maximum of 500ms for the operation to complete,
1401                  * if it's not complete by then we return error.
1402                 */
1403                         time = 0;
1404                         while (TRUE) {
1405                                 val64 = readq(&bar0->tti_command_mem);
1406                                 if (!(val64 & TTI_CMD_MEM_STROBE_NEW_CMD)) {
1407                                         break;
1408                                 }
1409                                 if (time > 10) {
1410                                         DBG_PRINT(ERR_DBG,
1411                                                 "%s: TTI init Failed\n",
1412                                         dev->name);
1413                                         return -1;
1414                                 }
1415                                 time++;
1416                                 msleep(50);
1417                         }
1418                 }
1419         } else {
1420
1421                 /* RTI Initialization */
1422                 if (nic->device_type == XFRAME_II_DEVICE) {
1423                         /*
1424                          * Programmed to generate Apprx 500 Intrs per
1425                          * second
1426                          */
1427                         int count = (nic->config.bus_speed * 125)/4;
1428                         val64 = RTI_DATA1_MEM_RX_TIMER_VAL(count);
1429                 } else {
1430                         val64 = RTI_DATA1_MEM_RX_TIMER_VAL(0xFFF);
1431                 }
1432                 val64 |= RTI_DATA1_MEM_RX_URNG_A(0xA) |
1433                     RTI_DATA1_MEM_RX_URNG_B(0x10) |
1434                     RTI_DATA1_MEM_RX_URNG_C(0x30) | RTI_DATA1_MEM_RX_TIMER_AC_EN;
1435
1436                 writeq(val64, &bar0->rti_data1_mem);
1437
1438                 val64 = RTI_DATA2_MEM_RX_UFC_A(0x1) |
1439                     RTI_DATA2_MEM_RX_UFC_B(0x2) ;
1440                 if (nic->intr_type == MSI_X)
1441                     val64 |= (RTI_DATA2_MEM_RX_UFC_C(0x20) | \
1442                                 RTI_DATA2_MEM_RX_UFC_D(0x40));
1443                 else
1444                     val64 |= (RTI_DATA2_MEM_RX_UFC_C(0x40) | \
1445                                 RTI_DATA2_MEM_RX_UFC_D(0x80));
1446                 writeq(val64, &bar0->rti_data2_mem);
1447
1448                 for (i = 0; i < config->rx_ring_num; i++) {
1449                         val64 = RTI_CMD_MEM_WE | RTI_CMD_MEM_STROBE_NEW_CMD
1450                                         | RTI_CMD_MEM_OFFSET(i);
1451                         writeq(val64, &bar0->rti_command_mem);
1452
1453                         /*
1454                          * Once the operation completes, the Strobe bit of the
1455                          * command register will be reset. We poll for this
1456                          * particular condition. We wait for a maximum of 500ms
1457                          * for the operation to complete, if it's not complete
1458                          * by then we return error.
1459                          */
1460                         time = 0;
1461                         while (TRUE) {
1462                                 val64 = readq(&bar0->rti_command_mem);
1463                                 if (!(val64 & RTI_CMD_MEM_STROBE_NEW_CMD)) {
1464                                         break;
1465                                 }
1466                                 if (time > 10) {
1467                                         DBG_PRINT(ERR_DBG, "%s: RTI init Failed\n",
1468                                                   dev->name);
1469                                         return -1;
1470                                 }
1471                                 time++;
1472                                 msleep(50);
1473                         }
1474                 }
1475         }
1476
1477         /*
1478          * Initializing proper values as Pause threshold into all
1479          * the 8 Queues on Rx side.
1480          */
1481         writeq(0xffbbffbbffbbffbbULL, &bar0->mc_pause_thresh_q0q3);
1482         writeq(0xffbbffbbffbbffbbULL, &bar0->mc_pause_thresh_q4q7);
1483
1484         /* Disable RMAC PAD STRIPPING */
1485         add = &bar0->mac_cfg;
1486         val64 = readq(&bar0->mac_cfg);
1487         val64 &= ~(MAC_CFG_RMAC_STRIP_PAD);
1488         writeq(RMAC_CFG_KEY(0x4C0D), &bar0->rmac_cfg_key);
1489         writel((u32) (val64), add);
1490         writeq(RMAC_CFG_KEY(0x4C0D), &bar0->rmac_cfg_key);
1491         writel((u32) (val64 >> 32), (add + 4));
1492         val64 = readq(&bar0->mac_cfg);
1493
1494         /* Enable FCS stripping by adapter */
1495         add = &bar0->mac_cfg;
1496         val64 = readq(&bar0->mac_cfg);
1497         val64 |= MAC_CFG_RMAC_STRIP_FCS;
1498         if (nic->device_type == XFRAME_II_DEVICE)
1499                 writeq(val64, &bar0->mac_cfg);
1500         else {
1501                 writeq(RMAC_CFG_KEY(0x4C0D), &bar0->rmac_cfg_key);
1502                 writel((u32) (val64), add);
1503                 writeq(RMAC_CFG_KEY(0x4C0D), &bar0->rmac_cfg_key);
1504                 writel((u32) (val64 >> 32), (add + 4));
1505         }
1506
1507         /*
1508          * Set the time value to be inserted in the pause frame
1509          * generated by xena.
1510          */
1511         val64 = readq(&bar0->rmac_pause_cfg);
1512         val64 &= ~(RMAC_PAUSE_HG_PTIME(0xffff));
1513         val64 |= RMAC_PAUSE_HG_PTIME(nic->mac_control.rmac_pause_time);
1514         writeq(val64, &bar0->rmac_pause_cfg);
1515
1516         /*
1517          * Set the Threshold Limit for Generating the pause frame
1518          * If the amount of data in any Queue exceeds ratio of
1519          * (mac_control.mc_pause_threshold_q0q3 or q4q7)/256
1520          * pause frame is generated
1521          */
1522         val64 = 0;
1523         for (i = 0; i < 4; i++) {
1524                 val64 |=
1525                     (((u64) 0xFF00 | nic->mac_control.
1526                       mc_pause_threshold_q0q3)
1527                      << (i * 2 * 8));
1528         }
1529         writeq(val64, &bar0->mc_pause_thresh_q0q3);
1530
1531         val64 = 0;
1532         for (i = 0; i < 4; i++) {
1533                 val64 |=
1534                     (((u64) 0xFF00 | nic->mac_control.
1535                       mc_pause_threshold_q4q7)
1536                      << (i * 2 * 8));
1537         }
1538         writeq(val64, &bar0->mc_pause_thresh_q4q7);
1539
1540         /*
1541          * TxDMA will stop Read request if the number of read split has
1542          * exceeded the limit pointed by shared_splits
1543          */
1544         val64 = readq(&bar0->pic_control);
1545         val64 |= PIC_CNTL_SHARED_SPLITS(shared_splits);
1546         writeq(val64, &bar0->pic_control);
1547
1548         /*
1549          * Programming the Herc to split every write transaction
1550          * that does not start on an ADB to reduce disconnects.
1551          */
1552         if (nic->device_type == XFRAME_II_DEVICE) {
1553                 val64 = WREQ_SPLIT_MASK_SET_MASK(255);
1554                 writeq(val64, &bar0->wreq_split_mask);
1555         }
1556
1557         /* Setting Link stability period to 64 ms */
1558         if (nic->device_type == XFRAME_II_DEVICE) {
1559                 val64 = MISC_LINK_STABILITY_PRD(3);
1560                 writeq(val64, &bar0->misc_control);
1561         }
1562
1563         return SUCCESS;
1564 }
1565 #define LINK_UP_DOWN_INTERRUPT          1
1566 #define MAC_RMAC_ERR_TIMER              2
1567
1568 static int s2io_link_fault_indication(nic_t *nic)
1569 {
1570         if (nic->intr_type != INTA)
1571                 return MAC_RMAC_ERR_TIMER;
1572         if (nic->device_type == XFRAME_II_DEVICE)
1573                 return LINK_UP_DOWN_INTERRUPT;
1574         else
1575                 return MAC_RMAC_ERR_TIMER;
1576 }
1577
1578 /**
1579  *  en_dis_able_nic_intrs - Enable or Disable the interrupts
1580  *  @nic: device private variable,
1581  *  @mask: A mask indicating which Intr block must be modified and,
1582  *  @flag: A flag indicating whether to enable or disable the Intrs.
1583  *  Description: This function will either disable or enable the interrupts
1584  *  depending on the flag argument. The mask argument can be used to
1585  *  enable/disable any Intr block.
1586  *  Return Value: NONE.
1587  */
1588
1589 static void en_dis_able_nic_intrs(struct s2io_nic *nic, u16 mask, int flag)
1590 {
1591         XENA_dev_config_t __iomem *bar0 = nic->bar0;
1592         register u64 val64 = 0, temp64 = 0;
1593
1594         /*  Top level interrupt classification */
1595         /*  PIC Interrupts */
1596         if ((mask & (TX_PIC_INTR | RX_PIC_INTR))) {
1597                 /*  Enable PIC Intrs in the general intr mask register */
1598                 val64 = TXPIC_INT_M | PIC_RX_INT_M;
1599                 if (flag == ENABLE_INTRS) {
1600                         temp64 = readq(&bar0->general_int_mask);
1601                         temp64 &= ~((u64) val64);
1602                         writeq(temp64, &bar0->general_int_mask);
1603                         /*
1604                          * If Hercules adapter enable GPIO otherwise
1605                          * disabled all PCIX, Flash, MDIO, IIC and GPIO
1606                          * interrupts for now.
1607                          * TODO
1608                          */
1609                         if (s2io_link_fault_indication(nic) ==
1610                                         LINK_UP_DOWN_INTERRUPT ) {
1611                                 temp64 = readq(&bar0->pic_int_mask);
1612                                 temp64 &= ~((u64) PIC_INT_GPIO);
1613                                 writeq(temp64, &bar0->pic_int_mask);
1614                                 temp64 = readq(&bar0->gpio_int_mask);
1615                                 temp64 &= ~((u64) GPIO_INT_MASK_LINK_UP);
1616                                 writeq(temp64, &bar0->gpio_int_mask);
1617                         } else {
1618                                 writeq(DISABLE_ALL_INTRS, &bar0->pic_int_mask);
1619                         }
1620                         /*
1621                          * No MSI Support is available presently, so TTI and
1622                          * RTI interrupts are also disabled.
1623                          */
1624                 } else if (flag == DISABLE_INTRS) {
1625                         /*
1626                          * Disable PIC Intrs in the general
1627                          * intr mask register
1628                          */
1629                         writeq(DISABLE_ALL_INTRS, &bar0->pic_int_mask);
1630                         temp64 = readq(&bar0->general_int_mask);
1631                         val64 |= temp64;
1632                         writeq(val64, &bar0->general_int_mask);
1633                 }
1634         }
1635
1636         /*  DMA Interrupts */
1637         /*  Enabling/Disabling Tx DMA interrupts */
1638         if (mask & TX_DMA_INTR) {
1639                 /* Enable TxDMA Intrs in the general intr mask register */
1640                 val64 = TXDMA_INT_M;
1641                 if (flag == ENABLE_INTRS) {
1642                         temp64 = readq(&bar0->general_int_mask);
1643                         temp64 &= ~((u64) val64);
1644                         writeq(temp64, &bar0->general_int_mask);
1645                         /*
1646                          * Keep all interrupts other than PFC interrupt
1647                          * and PCC interrupt disabled in DMA level.
1648                          */
1649                         val64 = DISABLE_ALL_INTRS & ~(TXDMA_PFC_INT_M |
1650                                                       TXDMA_PCC_INT_M);
1651                         writeq(val64, &bar0->txdma_int_mask);
1652                         /*
1653                          * Enable only the MISC error 1 interrupt in PFC block
1654                          */
1655                         val64 = DISABLE_ALL_INTRS & (~PFC_MISC_ERR_1);
1656                         writeq(val64, &bar0->pfc_err_mask);
1657                         /*
1658                          * Enable only the FB_ECC error interrupt in PCC block
1659                          */
1660                         val64 = DISABLE_ALL_INTRS & (~PCC_FB_ECC_ERR);
1661                         writeq(val64, &bar0->pcc_err_mask);
1662                 } else if (flag == DISABLE_INTRS) {
1663                         /*
1664                          * Disable TxDMA Intrs in the general intr mask
1665                          * register
1666                          */
1667                         writeq(DISABLE_ALL_INTRS, &bar0->txdma_int_mask);
1668                         writeq(DISABLE_ALL_INTRS, &bar0->pfc_err_mask);
1669                         temp64 = readq(&bar0->general_int_mask);
1670                         val64 |= temp64;
1671                         writeq(val64, &bar0->general_int_mask);
1672                 }
1673         }
1674
1675         /*  Enabling/Disabling Rx DMA interrupts */
1676         if (mask & RX_DMA_INTR) {
1677                 /*  Enable RxDMA Intrs in the general intr mask register */
1678                 val64 = RXDMA_INT_M;
1679                 if (flag == ENABLE_INTRS) {
1680                         temp64 = readq(&bar0->general_int_mask);
1681                         temp64 &= ~((u64) val64);
1682                         writeq(temp64, &bar0->general_int_mask);
1683                         /*
1684                          * All RxDMA block interrupts are disabled for now
1685                          * TODO
1686                          */
1687                         writeq(DISABLE_ALL_INTRS, &bar0->rxdma_int_mask);
1688                 } else if (flag == DISABLE_INTRS) {
1689                         /*
1690                          * Disable RxDMA Intrs in the general intr mask
1691                          * register
1692                          */
1693                         writeq(DISABLE_ALL_INTRS, &bar0->rxdma_int_mask);
1694                         temp64 = readq(&bar0->general_int_mask);
1695                         val64 |= temp64;
1696                         writeq(val64, &bar0->general_int_mask);
1697                 }
1698         }
1699
1700         /*  MAC Interrupts */
1701         /*  Enabling/Disabling MAC interrupts */
1702         if (mask & (TX_MAC_INTR | RX_MAC_INTR)) {
1703                 val64 = TXMAC_INT_M | RXMAC_INT_M;
1704                 if (flag == ENABLE_INTRS) {
1705                         temp64 = readq(&bar0->general_int_mask);
1706                         temp64 &= ~((u64) val64);
1707                         writeq(temp64, &bar0->general_int_mask);
1708                         /*
1709                          * All MAC block error interrupts are disabled for now
1710                          * TODO
1711                          */
1712                 } else if (flag == DISABLE_INTRS) {
1713                         /*
1714                          * Disable MAC Intrs in the general intr mask register
1715                          */
1716                         writeq(DISABLE_ALL_INTRS, &bar0->mac_int_mask);
1717                         writeq(DISABLE_ALL_INTRS,
1718                                &bar0->mac_rmac_err_mask);
1719
1720                         temp64 = readq(&bar0->general_int_mask);
1721                         val64 |= temp64;
1722                         writeq(val64, &bar0->general_int_mask);
1723                 }
1724         }
1725
1726         /*  XGXS Interrupts */
1727         if (mask & (TX_XGXS_INTR | RX_XGXS_INTR)) {
1728                 val64 = TXXGXS_INT_M | RXXGXS_INT_M;
1729                 if (flag == ENABLE_INTRS) {
1730                         temp64 = readq(&bar0->general_int_mask);
1731                         temp64 &= ~((u64) val64);
1732                         writeq(temp64, &bar0->general_int_mask);
1733                         /*
1734                          * All XGXS block error interrupts are disabled for now
1735                          * TODO
1736                          */
1737                         writeq(DISABLE_ALL_INTRS, &bar0->xgxs_int_mask);
1738                 } else if (flag == DISABLE_INTRS) {
1739                         /*
1740                          * Disable MC Intrs in the general intr mask register
1741                          */
1742                         writeq(DISABLE_ALL_INTRS, &bar0->xgxs_int_mask);
1743                         temp64 = readq(&bar0->general_int_mask);
1744                         val64 |= temp64;
1745                         writeq(val64, &bar0->general_int_mask);
1746                 }
1747         }
1748
1749         /*  Memory Controller(MC) interrupts */
1750         if (mask & MC_INTR) {
1751                 val64 = MC_INT_M;
1752                 if (flag == ENABLE_INTRS) {
1753                         temp64 = readq(&bar0->general_int_mask);
1754                         temp64 &= ~((u64) val64);
1755                         writeq(temp64, &bar0->general_int_mask);
1756                         /*
1757                          * Enable all MC Intrs.
1758                          */
1759                         writeq(0x0, &bar0->mc_int_mask);
1760                         writeq(0x0, &bar0->mc_err_mask);
1761                 } else if (flag == DISABLE_INTRS) {
1762                         /*
1763                          * Disable MC Intrs in the general intr mask register
1764                          */
1765                         writeq(DISABLE_ALL_INTRS, &bar0->mc_int_mask);
1766                         temp64 = readq(&bar0->general_int_mask);
1767                         val64 |= temp64;
1768                         writeq(val64, &bar0->general_int_mask);
1769                 }
1770         }
1771
1772
1773         /*  Tx traffic interrupts */
1774         if (mask & TX_TRAFFIC_INTR) {
1775                 val64 = TXTRAFFIC_INT_M;
1776                 if (flag == ENABLE_INTRS) {
1777                         temp64 = readq(&bar0->general_int_mask);
1778                         temp64 &= ~((u64) val64);
1779                         writeq(temp64, &bar0->general_int_mask);
1780                         /*
1781                          * Enable all the Tx side interrupts
1782                          * writing 0 Enables all 64 TX interrupt levels
1783                          */
1784                         writeq(0x0, &bar0->tx_traffic_mask);
1785                 } else if (flag == DISABLE_INTRS) {
1786                         /*
1787                          * Disable Tx Traffic Intrs in the general intr mask
1788                          * register.
1789                          */
1790                         writeq(DISABLE_ALL_INTRS, &bar0->tx_traffic_mask);
1791                         temp64 = readq(&bar0->general_int_mask);
1792                         val64 |= temp64;
1793                         writeq(val64, &bar0->general_int_mask);
1794                 }
1795         }
1796
1797         /*  Rx traffic interrupts */
1798         if (mask & RX_TRAFFIC_INTR) {
1799                 val64 = RXTRAFFIC_INT_M;
1800                 if (flag == ENABLE_INTRS) {
1801                         temp64 = readq(&bar0->general_int_mask);
1802                         temp64 &= ~((u64) val64);
1803                         writeq(temp64, &bar0->general_int_mask);
1804                         /* writing 0 Enables all 8 RX interrupt levels */
1805                         writeq(0x0, &bar0->rx_traffic_mask);
1806                 } else if (flag == DISABLE_INTRS) {
1807                         /*
1808                          * Disable Rx Traffic Intrs in the general intr mask
1809                          * register.
1810                          */
1811                         writeq(DISABLE_ALL_INTRS, &bar0->rx_traffic_mask);
1812                         temp64 = readq(&bar0->general_int_mask);
1813                         val64 |= temp64;
1814                         writeq(val64, &bar0->general_int_mask);
1815                 }
1816         }
1817 }
1818
1819 static int check_prc_pcc_state(u64 val64, int flag, int rev_id, int herc)
1820 {
1821         int ret = 0;
1822
1823         if (flag == FALSE) {
1824                 if ((!herc && (rev_id >= 4)) || herc) {
1825                         if (!(val64 & ADAPTER_STATUS_RMAC_PCC_IDLE) &&
1826                             ((val64 & ADAPTER_STATUS_RC_PRC_QUIESCENT) ==
1827                              ADAPTER_STATUS_RC_PRC_QUIESCENT)) {
1828                                 ret = 1;
1829                         }
1830                 }else {
1831                         if (!(val64 & ADAPTER_STATUS_RMAC_PCC_FOUR_IDLE) &&
1832                             ((val64 & ADAPTER_STATUS_RC_PRC_QUIESCENT) ==
1833                              ADAPTER_STATUS_RC_PRC_QUIESCENT)) {
1834                                 ret = 1;
1835                         }
1836                 }
1837         } else {
1838                 if ((!herc && (rev_id >= 4)) || herc) {
1839                         if (((val64 & ADAPTER_STATUS_RMAC_PCC_IDLE) ==
1840                              ADAPTER_STATUS_RMAC_PCC_IDLE) &&
1841                             (!(val64 & ADAPTER_STATUS_RC_PRC_QUIESCENT) ||
1842                              ((val64 & ADAPTER_STATUS_RC_PRC_QUIESCENT) ==
1843                               ADAPTER_STATUS_RC_PRC_QUIESCENT))) {
1844                                 ret = 1;
1845                         }
1846                 } else {
1847                         if (((val64 & ADAPTER_STATUS_RMAC_PCC_FOUR_IDLE) ==
1848                              ADAPTER_STATUS_RMAC_PCC_FOUR_IDLE) &&
1849                             (!(val64 & ADAPTER_STATUS_RC_PRC_QUIESCENT) ||
1850                              ((val64 & ADAPTER_STATUS_RC_PRC_QUIESCENT) ==
1851                               ADAPTER_STATUS_RC_PRC_QUIESCENT))) {
1852                                 ret = 1;
1853                         }
1854                 }
1855         }
1856
1857         return ret;
1858 }
1859 /**
1860  *  verify_xena_quiescence - Checks whether the H/W is ready
1861  *  @val64 :  Value read from adapter status register.
1862  *  @flag : indicates if the adapter enable bit was ever written once
1863  *  before.
1864  *  Description: Returns whether the H/W is ready to go or not. Depending
1865  *  on whether adapter enable bit was written or not the comparison
1866  *  differs and the calling function passes the input argument flag to
1867  *  indicate this.
1868  *  Return: 1 If xena is quiescence
1869  *          0 If Xena is not quiescence
1870  */
1871
1872 static int verify_xena_quiescence(nic_t *sp, u64 val64, int flag)
1873 {
1874         int ret = 0, herc;
1875         u64 tmp64 = ~((u64) val64);
1876         int rev_id = get_xena_rev_id(sp->pdev);
1877
1878         herc = (sp->device_type == XFRAME_II_DEVICE);
1879         if (!
1880             (tmp64 &
1881              (ADAPTER_STATUS_TDMA_READY | ADAPTER_STATUS_RDMA_READY |
1882               ADAPTER_STATUS_PFC_READY | ADAPTER_STATUS_TMAC_BUF_EMPTY |
1883               ADAPTER_STATUS_PIC_QUIESCENT | ADAPTER_STATUS_MC_DRAM_READY |
1884               ADAPTER_STATUS_MC_QUEUES_READY | ADAPTER_STATUS_M_PLL_LOCK |
1885               ADAPTER_STATUS_P_PLL_LOCK))) {
1886                 ret = check_prc_pcc_state(val64, flag, rev_id, herc);
1887         }
1888
1889         return ret;
1890 }
1891
1892 /**
1893  * fix_mac_address -  Fix for Mac addr problem on Alpha platforms
1894  * @sp: Pointer to device specifc structure
1895  * Description :
1896  * New procedure to clear mac address reading  problems on Alpha platforms
1897  *
1898  */
1899
1900 static void fix_mac_address(nic_t * sp)
1901 {
1902         XENA_dev_config_t __iomem *bar0 = sp->bar0;
1903         u64 val64;
1904         int i = 0;
1905
1906         while (fix_mac[i] != END_SIGN) {
1907                 writeq(fix_mac[i++], &bar0->gpio_control);
1908                 udelay(10);
1909                 val64 = readq(&bar0->gpio_control);
1910         }
1911 }
1912
1913 /**
1914  *  start_nic - Turns the device on
1915  *  @nic : device private variable.
1916  *  Description:
1917  *  This function actually turns the device on. Before this  function is
1918  *  called,all Registers are configured from their reset states
1919  *  and shared memory is allocated but the NIC is still quiescent. On
1920  *  calling this function, the device interrupts are cleared and the NIC is
1921  *  literally switched on by writing into the adapter control register.
1922  *  Return Value:
1923  *  SUCCESS on success and -1 on failure.
1924  */
1925
1926 static int start_nic(struct s2io_nic *nic)
1927 {
1928         XENA_dev_config_t __iomem *bar0 = nic->bar0;
1929         struct net_device *dev = nic->dev;
1930         register u64 val64 = 0;
1931         u16 interruptible;
1932         u16 subid, i;
1933         mac_info_t *mac_control;
1934         struct config_param *config;
1935
1936         mac_control = &nic->mac_control;
1937         config = &nic->config;
1938
1939         /*  PRC Initialization and configuration */
1940         for (i = 0; i < config->rx_ring_num; i++) {
1941                 writeq((u64) mac_control->rings[i].rx_blocks[0].block_dma_addr,
1942                        &bar0->prc_rxd0_n[i]);
1943
1944                 val64 = readq(&bar0->prc_ctrl_n[i]);
1945                 if (nic->config.bimodal)
1946                         val64 |= PRC_CTRL_BIMODAL_INTERRUPT;
1947                 if (nic->rxd_mode == RXD_MODE_1)
1948                         val64 |= PRC_CTRL_RC_ENABLED;
1949                 else
1950                         val64 |= PRC_CTRL_RC_ENABLED | PRC_CTRL_RING_MODE_3;
1951                 writeq(val64, &bar0->prc_ctrl_n[i]);
1952         }
1953
1954         if (nic->rxd_mode == RXD_MODE_3B) {
1955                 /* Enabling 2 buffer mode by writing into Rx_pa_cfg reg. */
1956                 val64 = readq(&bar0->rx_pa_cfg);
1957                 val64 |= RX_PA_CFG_IGNORE_L2_ERR;
1958                 writeq(val64, &bar0->rx_pa_cfg);
1959         }
1960
1961         /*
1962          * Enabling MC-RLDRAM. After enabling the device, we timeout
1963          * for around 100ms, which is approximately the time required
1964          * for the device to be ready for operation.
1965          */
1966         val64 = readq(&bar0->mc_rldram_mrs);
1967         val64 |= MC_RLDRAM_QUEUE_SIZE_ENABLE | MC_RLDRAM_MRS_ENABLE;
1968         SPECIAL_REG_WRITE(val64, &bar0->mc_rldram_mrs, UF);
1969         val64 = readq(&bar0->mc_rldram_mrs);
1970
1971         msleep(100);    /* Delay by around 100 ms. */
1972
1973         /* Enabling ECC Protection. */
1974         val64 = readq(&bar0->adapter_control);
1975         val64 &= ~ADAPTER_ECC_EN;
1976         writeq(val64, &bar0->adapter_control);
1977
1978         /*
1979          * Clearing any possible Link state change interrupts that
1980          * could have popped up just before Enabling the card.
1981          */
1982         val64 = readq(&bar0->mac_rmac_err_reg);
1983         if (val64)
1984                 writeq(val64, &bar0->mac_rmac_err_reg);
1985
1986         /*
1987          * Verify if the device is ready to be enabled, if so enable
1988          * it.
1989          */
1990         val64 = readq(&bar0->adapter_status);
1991         if (!verify_xena_quiescence(nic, val64, nic->device_enabled_once)) {
1992                 DBG_PRINT(ERR_DBG, "%s: device is not ready, ", dev->name);
1993                 DBG_PRINT(ERR_DBG, "Adapter status reads: 0x%llx\n",
1994                           (unsigned long long) val64);
1995                 return FAILURE;
1996         }
1997
1998         /*  Enable select interrupts */
1999         if (nic->intr_type != INTA)
2000                 en_dis_able_nic_intrs(nic, ENA_ALL_INTRS, DISABLE_INTRS);
2001         else {
2002                 interruptible = TX_TRAFFIC_INTR | RX_TRAFFIC_INTR;
2003                 interruptible |= TX_PIC_INTR | RX_PIC_INTR;
2004                 interruptible |= TX_MAC_INTR | RX_MAC_INTR;
2005                 en_dis_able_nic_intrs(nic, interruptible, ENABLE_INTRS);
2006         }
2007
2008         /*
2009          * With some switches, link might be already up at this point.
2010          * Because of this weird behavior, when we enable laser,
2011          * we may not get link. We need to handle this. We cannot
2012          * figure out which switch is misbehaving. So we are forced to
2013          * make a global change.
2014          */
2015
2016         /* Enabling Laser. */
2017         val64 = readq(&bar0->adapter_control);
2018         val64 |= ADAPTER_EOI_TX_ON;
2019         writeq(val64, &bar0->adapter_control);
2020
2021         /* SXE-002: Initialize link and activity LED */
2022         subid = nic->pdev->subsystem_device;
2023         if (((subid & 0xFF) >= 0x07) &&
2024             (nic->device_type == XFRAME_I_DEVICE)) {
2025                 val64 = readq(&bar0->gpio_control);
2026                 val64 |= 0x0000800000000000ULL;
2027                 writeq(val64, &bar0->gpio_control);
2028                 val64 = 0x0411040400000000ULL;
2029                 writeq(val64, (void __iomem *)bar0 + 0x2700);
2030         }
2031
2032         /*
2033          * Don't see link state interrupts on certain switches, so
2034          * directly scheduling a link state task from here.
2035          */
2036         schedule_work(&nic->set_link_task);
2037
2038         return SUCCESS;
2039 }
2040 /**
2041  * s2io_txdl_getskb - Get the skb from txdl, unmap and return skb
2042  */
2043 static struct sk_buff *s2io_txdl_getskb(fifo_info_t *fifo_data, TxD_t *txdlp, int get_off)
2044 {
2045         nic_t *nic = fifo_data->nic;
2046         struct sk_buff *skb;
2047         TxD_t *txds;
2048         u16 j, frg_cnt;
2049
2050         txds = txdlp;
2051         if (txds->Host_Control == (u64)(long)nic->ufo_in_band_v) {
2052                 pci_unmap_single(nic->pdev, (dma_addr_t)
2053                         txds->Buffer_Pointer, sizeof(u64),
2054                         PCI_DMA_TODEVICE);
2055                 txds++;
2056         }
2057
2058         skb = (struct sk_buff *) ((unsigned long)
2059                         txds->Host_Control);
2060         if (!skb) {
2061                 memset(txdlp, 0, (sizeof(TxD_t) * fifo_data->max_txds));
2062                 return NULL;
2063         }
2064         pci_unmap_single(nic->pdev, (dma_addr_t)
2065                          txds->Buffer_Pointer,
2066                          skb->len - skb->data_len,
2067                          PCI_DMA_TODEVICE);
2068         frg_cnt = skb_shinfo(skb)->nr_frags;
2069         if (frg_cnt) {
2070                 txds++;
2071                 for (j = 0; j < frg_cnt; j++, txds++) {
2072                         skb_frag_t *frag = &skb_shinfo(skb)->frags[j];
2073                         if (!txds->Buffer_Pointer)
2074                                 break;
2075                         pci_unmap_page(nic->pdev, (dma_addr_t)
2076                                         txds->Buffer_Pointer,
2077                                        frag->size, PCI_DMA_TODEVICE);
2078                 }
2079         }
2080         txdlp->Host_Control = 0;
2081         return(skb);
2082 }
2083
2084 /**
2085  *  free_tx_buffers - Free all queued Tx buffers
2086  *  @nic : device private variable.
2087  *  Description:
2088  *  Free all queued Tx buffers.
2089  *  Return Value: void
2090 */
2091
2092 static void free_tx_buffers(struct s2io_nic *nic)
2093 {
2094         struct net_device *dev = nic->dev;
2095         struct sk_buff *skb;
2096         TxD_t *txdp;
2097         int i, j;
2098         mac_info_t *mac_control;
2099         struct config_param *config;
2100         int cnt = 0;
2101
2102         mac_control = &nic->mac_control;
2103         config = &nic->config;
2104
2105         for (i = 0; i < config->tx_fifo_num; i++) {
2106                 for (j = 0; j < config->tx_cfg[i].fifo_len - 1; j++) {
2107                         txdp = (TxD_t *) mac_control->fifos[i].list_info[j].
2108                             list_virt_addr;
2109                         skb = s2io_txdl_getskb(&mac_control->fifos[i], txdp, j);
2110                         if (skb) {
2111                                 dev_kfree_skb(skb);
2112                                 cnt++;
2113                         }
2114                 }
2115                 DBG_PRINT(INTR_DBG,
2116                           "%s:forcibly freeing %d skbs on FIFO%d\n",
2117                           dev->name, cnt, i);
2118                 mac_control->fifos[i].tx_curr_get_info.offset = 0;
2119                 mac_control->fifos[i].tx_curr_put_info.offset = 0;
2120         }
2121 }
2122
2123 /**
2124  *   stop_nic -  To stop the nic
2125  *   @nic ; device private variable.
2126  *   Description:
2127  *   This function does exactly the opposite of what the start_nic()
2128  *   function does. This function is called to stop the device.
2129  *   Return Value:
2130  *   void.
2131  */
2132
2133 static void stop_nic(struct s2io_nic *nic)
2134 {
2135         XENA_dev_config_t __iomem *bar0 = nic->bar0;
2136         register u64 val64 = 0;
2137         u16 interruptible, i;
2138         mac_info_t *mac_control;
2139         struct config_param *config;
2140
2141         mac_control = &nic->mac_control;
2142         config = &nic->config;
2143
2144         /*  Disable all interrupts */
2145         interruptible = TX_TRAFFIC_INTR | RX_TRAFFIC_INTR;
2146         interruptible |= TX_PIC_INTR | RX_PIC_INTR;
2147         interruptible |= TX_MAC_INTR | RX_MAC_INTR;
2148         en_dis_able_nic_intrs(nic, interruptible, DISABLE_INTRS);
2149
2150         /*  Disable PRCs */
2151         for (i = 0; i < config->rx_ring_num; i++) {
2152                 val64 = readq(&bar0->prc_ctrl_n[i]);
2153                 val64 &= ~((u64) PRC_CTRL_RC_ENABLED);
2154                 writeq(val64, &bar0->prc_ctrl_n[i]);
2155         }
2156 }
2157
2158 static int fill_rxd_3buf(nic_t *nic, RxD_t *rxdp, struct sk_buff *skb)
2159 {
2160         struct net_device *dev = nic->dev;
2161         struct sk_buff *frag_list;
2162         void *tmp;
2163
2164         /* Buffer-1 receives L3/L4 headers */
2165         ((RxD3_t*)rxdp)->Buffer1_ptr = pci_map_single
2166                         (nic->pdev, skb->data, l3l4hdr_size + 4,
2167                         PCI_DMA_FROMDEVICE);
2168
2169         /* skb_shinfo(skb)->frag_list will have L4 data payload */
2170         skb_shinfo(skb)->frag_list = dev_alloc_skb(dev->mtu + ALIGN_SIZE);
2171         if (skb_shinfo(skb)->frag_list == NULL) {
2172                 DBG_PRINT(ERR_DBG, "%s: dev_alloc_skb failed\n ", dev->name);
2173                 return -ENOMEM ;
2174         }
2175         frag_list = skb_shinfo(skb)->frag_list;
2176         frag_list->next = NULL;
2177         tmp = (void *)ALIGN((long)frag_list->data, ALIGN_SIZE + 1);
2178         frag_list->data = tmp;
2179         frag_list->tail = tmp;
2180
2181         /* Buffer-2 receives L4 data payload */
2182         ((RxD3_t*)rxdp)->Buffer2_ptr = pci_map_single(nic->pdev,
2183                                 frag_list->data, dev->mtu,
2184                                 PCI_DMA_FROMDEVICE);
2185         rxdp->Control_2 |= SET_BUFFER1_SIZE_3(l3l4hdr_size + 4);
2186         rxdp->Control_2 |= SET_BUFFER2_SIZE_3(dev->mtu);
2187
2188         return SUCCESS;
2189 }
2190
2191 /**
2192  *  fill_rx_buffers - Allocates the Rx side skbs
2193  *  @nic:  device private variable
2194  *  @ring_no: ring number
2195  *  Description:
2196  *  The function allocates Rx side skbs and puts the physical
2197  *  address of these buffers into the RxD buffer pointers, so that the NIC
2198  *  can DMA the received frame into these locations.
2199  *  The NIC supports 3 receive modes, viz
2200  *  1. single buffer,
2201  *  2. three buffer and
2202  *  3. Five buffer modes.
2203  *  Each mode defines how many fragments the received frame will be split
2204  *  up into by the NIC. The frame is split into L3 header, L4 Header,
2205  *  L4 payload in three buffer mode and in 5 buffer mode, L4 payload itself
2206  *  is split into 3 fragments. As of now only single buffer mode is
2207  *  supported.
2208  *   Return Value:
2209  *  SUCCESS on success or an appropriate -ve value on failure.
2210  */
2211
2212 static int fill_rx_buffers(struct s2io_nic *nic, int ring_no)
2213 {
2214         struct net_device *dev = nic->dev;
2215         struct sk_buff *skb;
2216         RxD_t *rxdp;
2217         int off, off1, size, block_no, block_no1;
2218         u32 alloc_tab = 0;
2219         u32 alloc_cnt;
2220         mac_info_t *mac_control;
2221         struct config_param *config;
2222         u64 tmp;
2223         buffAdd_t *ba;
2224 #ifndef CONFIG_S2IO_NAPI
2225         unsigned long flags;
2226 #endif
2227         RxD_t *first_rxdp = NULL;
2228
2229         mac_control = &nic->mac_control;
2230         config = &nic->config;
2231         alloc_cnt = mac_control->rings[ring_no].pkt_cnt -
2232             atomic_read(&nic->rx_bufs_left[ring_no]);
2233
2234         while (alloc_tab < alloc_cnt) {
2235                 block_no = mac_control->rings[ring_no].rx_curr_put_info.
2236                     block_index;
2237                 block_no1 = mac_control->rings[ring_no].rx_curr_get_info.
2238                     block_index;
2239                 off = mac_control->rings[ring_no].rx_curr_put_info.offset;
2240                 off1 = mac_control->rings[ring_no].rx_curr_get_info.offset;
2241
2242                 rxdp = mac_control->rings[ring_no].
2243                                 rx_blocks[block_no].rxds[off].virt_addr;
2244
2245                 if ((block_no == block_no1) && (off == off1) &&
2246                                         (rxdp->Host_Control)) {
2247                         DBG_PRINT(INTR_DBG, "%s: Get and Put",
2248                                   dev->name);
2249                         DBG_PRINT(INTR_DBG, " info equated\n");
2250                         goto end;
2251                 }
2252                 if (off && (off == rxd_count[nic->rxd_mode])) {
2253                         mac_control->rings[ring_no].rx_curr_put_info.
2254                             block_index++;
2255                         if (mac_control->rings[ring_no].rx_curr_put_info.
2256                             block_index == mac_control->rings[ring_no].
2257                                         block_count)
2258                                 mac_control->rings[ring_no].rx_curr_put_info.
2259                                         block_index = 0;
2260                         block_no = mac_control->rings[ring_no].
2261                                         rx_curr_put_info.block_index;
2262                         if (off == rxd_count[nic->rxd_mode])
2263                                 off = 0;
2264                         mac_control->rings[ring_no].rx_curr_put_info.
2265                                 offset = off;
2266                         rxdp = mac_control->rings[ring_no].
2267                                 rx_blocks[block_no].block_virt_addr;
2268                         DBG_PRINT(INTR_DBG, "%s: Next block at: %p\n",
2269                                   dev->name, rxdp);
2270                 }
2271 #ifndef CONFIG_S2IO_NAPI
2272                 spin_lock_irqsave(&nic->put_lock, flags);
2273                 mac_control->rings[ring_no].put_pos =
2274                     (block_no * (rxd_count[nic->rxd_mode] + 1)) + off;
2275                 spin_unlock_irqrestore(&nic->put_lock, flags);
2276 #endif
2277                 if ((rxdp->Control_1 & RXD_OWN_XENA) &&
2278                         ((nic->rxd_mode >= RXD_MODE_3A) &&
2279                                 (rxdp->Control_2 & BIT(0)))) {
2280                         mac_control->rings[ring_no].rx_curr_put_info.
2281                                         offset = off;
2282                         goto end;
2283                 }
2284                 /* calculate size of skb based on ring mode */
2285                 size = dev->mtu + HEADER_ETHERNET_II_802_3_SIZE +
2286                                 HEADER_802_2_SIZE + HEADER_SNAP_SIZE;
2287                 if (nic->rxd_mode == RXD_MODE_1)
2288                         size += NET_IP_ALIGN;
2289                 else if (nic->rxd_mode == RXD_MODE_3B)
2290                         size = dev->mtu + ALIGN_SIZE + BUF0_LEN + 4;
2291                 else
2292                         size = l3l4hdr_size + ALIGN_SIZE + BUF0_LEN + 4;
2293
2294                 /* allocate skb */
2295                 skb = dev_alloc_skb(size);
2296                 if(!skb) {
2297                         DBG_PRINT(ERR_DBG, "%s: Out of ", dev->name);
2298                         DBG_PRINT(ERR_DBG, "memory to allocate SKBs\n");
2299                         if (first_rxdp) {
2300                                 wmb();
2301                                 first_rxdp->Control_1 |= RXD_OWN_XENA;
2302                         }
2303                         return -ENOMEM ;
2304                 }
2305                 if (nic->rxd_mode == RXD_MODE_1) {
2306                         /* 1 buffer mode - normal operation mode */
2307                         memset(rxdp, 0, sizeof(RxD1_t));
2308                         skb_reserve(skb, NET_IP_ALIGN);
2309                         ((RxD1_t*)rxdp)->Buffer0_ptr = pci_map_single
2310                             (nic->pdev, skb->data, size, PCI_DMA_FROMDEVICE);
2311                         rxdp->Control_2 &= (~MASK_BUFFER0_SIZE_1);
2312                         rxdp->Control_2 |= SET_BUFFER0_SIZE_1(size);
2313
2314                 } else if (nic->rxd_mode >= RXD_MODE_3A) {
2315                         /*
2316                          * 2 or 3 buffer mode -
2317                          * Both 2 buffer mode and 3 buffer mode provides 128
2318                          * byte aligned receive buffers.
2319                          *
2320                          * 3 buffer mode provides header separation where in
2321                          * skb->data will have L3/L4 headers where as
2322                          * skb_shinfo(skb)->frag_list will have the L4 data
2323                          * payload
2324                          */
2325
2326                         memset(rxdp, 0, sizeof(RxD3_t));
2327                         ba = &mac_control->rings[ring_no].ba[block_no][off];
2328                         skb_reserve(skb, BUF0_LEN);
2329                         tmp = (u64)(unsigned long) skb->data;
2330                         tmp += ALIGN_SIZE;
2331                         tmp &= ~ALIGN_SIZE;
2332                         skb->data = (void *) (unsigned long)tmp;
2333                         skb->tail = (void *) (unsigned long)tmp;
2334
2335                         ((RxD3_t*)rxdp)->Buffer0_ptr =
2336                             pci_map_single(nic->pdev, ba->ba_0, BUF0_LEN,
2337                                            PCI_DMA_FROMDEVICE);
2338                         rxdp->Control_2 = SET_BUFFER0_SIZE_3(BUF0_LEN);
2339                         if (nic->rxd_mode == RXD_MODE_3B) {
2340                                 /* Two buffer mode */
2341
2342                                 /*
2343                                  * Buffer2 will have L3/L4 header plus
2344                                  * L4 payload
2345                                  */
2346                                 ((RxD3_t*)rxdp)->Buffer2_ptr = pci_map_single
2347                                 (nic->pdev, skb->data, dev->mtu + 4,
2348                                                 PCI_DMA_FROMDEVICE);
2349
2350                                 /* Buffer-1 will be dummy buffer not used */
2351                                 ((RxD3_t*)rxdp)->Buffer1_ptr =
2352                                 pci_map_single(nic->pdev, ba->ba_1, BUF1_LEN,
2353                                         PCI_DMA_FROMDEVICE);
2354                                 rxdp->Control_2 |= SET_BUFFER1_SIZE_3(1);
2355                                 rxdp->Control_2 |= SET_BUFFER2_SIZE_3
2356                                                                 (dev->mtu + 4);
2357                         } else {
2358                                 /* 3 buffer mode */
2359                                 if (fill_rxd_3buf(nic, rxdp, skb) == -ENOMEM) {
2360                                         dev_kfree_skb_irq(skb);
2361                                         if (first_rxdp) {
2362                                                 wmb();
2363                                                 first_rxdp->Control_1 |=
2364                                                         RXD_OWN_XENA;
2365                                         }
2366                                         return -ENOMEM ;
2367                                 }
2368                         }
2369                         rxdp->Control_2 |= BIT(0);
2370                 }
2371                 rxdp->Host_Control = (unsigned long) (skb);
2372                 if (alloc_tab & ((1 << rxsync_frequency) - 1))
2373                         rxdp->Control_1 |= RXD_OWN_XENA;
2374                 off++;
2375                 if (off == (rxd_count[nic->rxd_mode] + 1))
2376                         off = 0;
2377                 mac_control->rings[ring_no].rx_curr_put_info.offset = off;
2378
2379                 rxdp->Control_2 |= SET_RXD_MARKER;
2380                 if (!(alloc_tab & ((1 << rxsync_frequency) - 1))) {
2381                         if (first_rxdp) {
2382                                 wmb();
2383                                 first_rxdp->Control_1 |= RXD_OWN_XENA;
2384                         }
2385                         first_rxdp = rxdp;
2386                 }
2387                 atomic_inc(&nic->rx_bufs_left[ring_no]);
2388                 alloc_tab++;
2389         }
2390
2391       end:
2392         /* Transfer ownership of first descriptor to adapter just before
2393          * exiting. Before that, use memory barrier so that ownership
2394          * and other fields are seen by adapter correctly.
2395          */
2396         if (first_rxdp) {
2397                 wmb();
2398                 first_rxdp->Control_1 |= RXD_OWN_XENA;
2399         }
2400
2401         return SUCCESS;
2402 }
2403
2404 static void free_rxd_blk(struct s2io_nic *sp, int ring_no, int blk)
2405 {
2406         struct net_device *dev = sp->dev;
2407         int j;
2408         struct sk_buff *skb;
2409         RxD_t *rxdp;
2410         mac_info_t *mac_control;
2411         buffAdd_t *ba;
2412
2413         mac_control = &sp->mac_control;
2414         for (j = 0 ; j < rxd_count[sp->rxd_mode]; j++) {
2415                 rxdp = mac_control->rings[ring_no].
2416                                 rx_blocks[blk].rxds[j].virt_addr;
2417                 skb = (struct sk_buff *)
2418                         ((unsigned long) rxdp->Host_Control);
2419                 if (!skb) {
2420                         continue;
2421                 }
2422                 if (sp->rxd_mode == RXD_MODE_1) {
2423                         pci_unmap_single(sp->pdev, (dma_addr_t)
2424                                  ((RxD1_t*)rxdp)->Buffer0_ptr,
2425                                  dev->mtu +
2426                                  HEADER_ETHERNET_II_802_3_SIZE
2427                                  + HEADER_802_2_SIZE +
2428                                  HEADER_SNAP_SIZE,
2429                                  PCI_DMA_FROMDEVICE);
2430                         memset(rxdp, 0, sizeof(RxD1_t));
2431                 } else if(sp->rxd_mode == RXD_MODE_3B) {
2432                         ba = &mac_control->rings[ring_no].
2433                                 ba[blk][j];
2434                         pci_unmap_single(sp->pdev, (dma_addr_t)
2435                                  ((RxD3_t*)rxdp)->Buffer0_ptr,
2436                                  BUF0_LEN,
2437                                  PCI_DMA_FROMDEVICE);
2438                         pci_unmap_single(sp->pdev, (dma_addr_t)
2439                                  ((RxD3_t*)rxdp)->Buffer1_ptr,
2440                                  BUF1_LEN,
2441                                  PCI_DMA_FROMDEVICE);
2442                         pci_unmap_single(sp->pdev, (dma_addr_t)
2443                                  ((RxD3_t*)rxdp)->Buffer2_ptr,
2444                                  dev->mtu + 4,
2445                                  PCI_DMA_FROMDEVICE);
2446                         memset(rxdp, 0, sizeof(RxD3_t));
2447                 } else {
2448                         pci_unmap_single(sp->pdev, (dma_addr_t)
2449                                 ((RxD3_t*)rxdp)->Buffer0_ptr, BUF0_LEN,
2450                                 PCI_DMA_FROMDEVICE);
2451                         pci_unmap_single(sp->pdev, (dma_addr_t)
2452                                 ((RxD3_t*)rxdp)->Buffer1_ptr,
2453                                 l3l4hdr_size + 4,
2454                                 PCI_DMA_FROMDEVICE);
2455                         pci_unmap_single(sp->pdev, (dma_addr_t)
2456                                 ((RxD3_t*)rxdp)->Buffer2_ptr, dev->mtu,
2457                                 PCI_DMA_FROMDEVICE);
2458                         memset(rxdp, 0, sizeof(RxD3_t));
2459                 }
2460                 dev_kfree_skb(skb);
2461                 atomic_dec(&sp->rx_bufs_left[ring_no]);
2462         }
2463 }
2464
2465 /**
2466  *  free_rx_buffers - Frees all Rx buffers
2467  *  @sp: device private variable.
2468  *  Description:
2469  *  This function will free all Rx buffers allocated by host.
2470  *  Return Value:
2471  *  NONE.
2472  */
2473
2474 static void free_rx_buffers(struct s2io_nic *sp)
2475 {
2476         struct net_device *dev = sp->dev;
2477         int i, blk = 0, buf_cnt = 0;
2478         mac_info_t *mac_control;
2479         struct config_param *config;
2480
2481         mac_control = &sp->mac_control;
2482         config = &sp->config;
2483
2484         for (i = 0; i < config->rx_ring_num; i++) {
2485                 for (blk = 0; blk < rx_ring_sz[i]; blk++)
2486                         free_rxd_blk(sp,i,blk);
2487
2488                 mac_control->rings[i].rx_curr_put_info.block_index = 0;
2489                 mac_control->rings[i].rx_curr_get_info.block_index = 0;
2490                 mac_control->rings[i].rx_curr_put_info.offset = 0;
2491                 mac_control->rings[i].rx_curr_get_info.offset = 0;
2492                 atomic_set(&sp->rx_bufs_left[i], 0);
2493                 DBG_PRINT(INIT_DBG, "%s:Freed 0x%x Rx Buffers on ring%d\n",
2494                           dev->name, buf_cnt, i);
2495         }
2496 }
2497
2498 /**
2499  * s2io_poll - Rx interrupt handler for NAPI support
2500  * @dev : pointer to the device structure.
2501  * @budget : The number of packets that were budgeted to be processed
2502  * during  one pass through the 'Poll" function.
2503  * Description:
2504  * Comes into picture only if NAPI support has been incorporated. It does
2505  * the same thing that rx_intr_handler does, but not in a interrupt context
2506  * also It will process only a given number of packets.
2507  * Return value:
2508  * 0 on success and 1 if there are No Rx packets to be processed.
2509  */
2510
2511 #if defined(CONFIG_S2IO_NAPI)
2512 static int s2io_poll(struct net_device *dev, int *budget)
2513 {
2514         nic_t *nic = dev->priv;
2515         int pkt_cnt = 0, org_pkts_to_process;
2516         mac_info_t *mac_control;
2517         struct config_param *config;
2518         XENA_dev_config_t __iomem *bar0 = nic->bar0;
2519         u64 val64;
2520         int i;
2521
2522         atomic_inc(&nic->isr_cnt);
2523         mac_control = &nic->mac_control;
2524         config = &nic->config;
2525
2526         nic->pkts_to_process = *budget;
2527         if (nic->pkts_to_process > dev->quota)
2528                 nic->pkts_to_process = dev->quota;
2529         org_pkts_to_process = nic->pkts_to_process;
2530
2531         val64 = readq(&bar0->rx_traffic_int);
2532         writeq(val64, &bar0->rx_traffic_int);
2533
2534         for (i = 0; i < config->rx_ring_num; i++) {
2535                 rx_intr_handler(&mac_control->rings[i]);
2536                 pkt_cnt = org_pkts_to_process - nic->pkts_to_process;
2537                 if (!nic->pkts_to_process) {
2538                         /* Quota for the current iteration has been met */
2539                         goto no_rx;
2540                 }
2541         }
2542         if (!pkt_cnt)
2543                 pkt_cnt = 1;
2544
2545         dev->quota -= pkt_cnt;
2546         *budget -= pkt_cnt;
2547         netif_rx_complete(dev);
2548
2549         for (i = 0; i < config->rx_ring_num; i++) {
2550                 if (fill_rx_buffers(nic, i) == -ENOMEM) {
2551                         DBG_PRINT(ERR_DBG, "%s:Out of memory", dev->name);
2552                         DBG_PRINT(ERR_DBG, " in Rx Poll!!\n");
2553                         break;
2554                 }
2555         }
2556         /* Re enable the Rx interrupts. */
2557         en_dis_able_nic_intrs(nic, RX_TRAFFIC_INTR, ENABLE_INTRS);
2558         atomic_dec(&nic->isr_cnt);
2559         return 0;
2560
2561 no_rx:
2562         dev->quota -= pkt_cnt;
2563         *budget -= pkt_cnt;
2564
2565         for (i = 0; i < config->rx_ring_num; i++) {
2566                 if (fill_rx_buffers(nic, i) == -ENOMEM) {
2567                         DBG_PRINT(ERR_DBG, "%s:Out of memory", dev->name);
2568                         DBG_PRINT(ERR_DBG, " in Rx Poll!!\n");
2569                         break;
2570                 }
2571         }
2572         atomic_dec(&nic->isr_cnt);
2573         return 1;
2574 }
2575 #endif
2576
2577 /**
2578  *  rx_intr_handler - Rx interrupt handler
2579  *  @nic: device private variable.
2580  *  Description:
2581  *  If the interrupt is because of a received frame or if the
2582  *  receive ring contains fresh as yet un-processed frames,this function is
2583  *  called. It picks out the RxD at which place the last Rx processing had
2584  *  stopped and sends the skb to the OSM's Rx handler and then increments
2585  *  the offset.
2586  *  Return Value:
2587  *  NONE.
2588  */
2589 static void rx_intr_handler(ring_info_t *ring_data)
2590 {
2591         nic_t *nic = ring_data->nic;
2592         struct net_device *dev = (struct net_device *) nic->dev;
2593         int get_block, put_block, put_offset;
2594         rx_curr_get_info_t get_info, put_info;
2595         RxD_t *rxdp;
2596         struct sk_buff *skb;
2597 #ifndef CONFIG_S2IO_NAPI
2598         int pkt_cnt = 0;
2599 #endif
2600         int i;
2601
2602         spin_lock(&nic->rx_lock);
2603         if (atomic_read(&nic->card_state) == CARD_DOWN) {
2604                 DBG_PRINT(INTR_DBG, "%s: %s going down for reset\n",
2605                           __FUNCTION__, dev->name);
2606                 spin_unlock(&nic->rx_lock);
2607                 return;
2608         }
2609
2610         get_info = ring_data->rx_curr_get_info;
2611         get_block = get_info.block_index;
2612         put_info = ring_data->rx_curr_put_info;
2613         put_block = put_info.block_index;
2614         rxdp = ring_data->rx_blocks[get_block].rxds[get_info.offset].virt_addr;
2615 #ifndef CONFIG_S2IO_NAPI
2616         spin_lock(&nic->put_lock);
2617         put_offset = ring_data->put_pos;
2618         spin_unlock(&nic->put_lock);
2619 #else
2620         put_offset = (put_block * (rxd_count[nic->rxd_mode] + 1)) +
2621                 put_info.offset;
2622 #endif
2623         while (RXD_IS_UP2DT(rxdp)) {
2624                 /* If your are next to put index then it's FIFO full condition */
2625                 if ((get_block == put_block) &&
2626                     (get_info.offset + 1) == put_info.offset) {
2627                         DBG_PRINT(ERR_DBG, "%s: Ring Full\n",dev->name);
2628                         break;
2629                 }
2630                 skb = (struct sk_buff *) ((unsigned long)rxdp->Host_Control);
2631                 if (skb == NULL) {
2632                         DBG_PRINT(ERR_DBG, "%s: The skb is ",
2633                                   dev->name);
2634                         DBG_PRINT(ERR_DBG, "Null in Rx Intr\n");
2635                         spin_unlock(&nic->rx_lock);
2636                         return;
2637                 }
2638                 if (nic->rxd_mode == RXD_MODE_1) {
2639                         pci_unmap_single(nic->pdev, (dma_addr_t)
2640                                  ((RxD1_t*)rxdp)->Buffer0_ptr,
2641                                  dev->mtu +
2642                                  HEADER_ETHERNET_II_802_3_SIZE +
2643                                  HEADER_802_2_SIZE +
2644                                  HEADER_SNAP_SIZE,
2645                                  PCI_DMA_FROMDEVICE);
2646                 } else if (nic->rxd_mode == RXD_MODE_3B) {
2647                         pci_unmap_single(nic->pdev, (dma_addr_t)
2648                                  ((RxD3_t*)rxdp)->Buffer0_ptr,
2649                                  BUF0_LEN, PCI_DMA_FROMDEVICE);
2650                         pci_unmap_single(nic->pdev, (dma_addr_t)
2651                                  ((RxD3_t*)rxdp)->Buffer1_ptr,
2652                                  BUF1_LEN, PCI_DMA_FROMDEVICE);
2653                         pci_unmap_single(nic->pdev, (dma_addr_t)
2654                                  ((RxD3_t*)rxdp)->Buffer2_ptr,
2655                                  dev->mtu + 4,
2656                                  PCI_DMA_FROMDEVICE);
2657                 } else {
2658                         pci_unmap_single(nic->pdev, (dma_addr_t)
2659                                          ((RxD3_t*)rxdp)->Buffer0_ptr, BUF0_LEN,
2660                                          PCI_DMA_FROMDEVICE);
2661                         pci_unmap_single(nic->pdev, (dma_addr_t)
2662                                          ((RxD3_t*)rxdp)->Buffer1_ptr,
2663                                          l3l4hdr_size + 4,
2664                                          PCI_DMA_FROMDEVICE);
2665                         pci_unmap_single(nic->pdev, (dma_addr_t)
2666                                          ((RxD3_t*)rxdp)->Buffer2_ptr,
2667                                          dev->mtu, PCI_DMA_FROMDEVICE);
2668                 }
2669                 rx_osm_handler(ring_data, rxdp);
2670                 get_info.offset++;
2671                 ring_data->rx_curr_get_info.offset = get_info.offset;
2672                 rxdp = ring_data->rx_blocks[get_block].
2673                                 rxds[get_info.offset].virt_addr;
2674                 if (get_info.offset == rxd_count[nic->rxd_mode]) {
2675                         get_info.offset = 0;
2676                         ring_data->rx_curr_get_info.offset = get_info.offset;
2677                         get_block++;
2678                         if (get_block == ring_data->block_count)
2679                                 get_block = 0;
2680                         ring_data->rx_curr_get_info.block_index = get_block;
2681                         rxdp = ring_data->rx_blocks[get_block].block_virt_addr;
2682                 }
2683
2684 #ifdef CONFIG_S2IO_NAPI
2685                 nic->pkts_to_process -= 1;
2686                 if (!nic->pkts_to_process)
2687                         break;
2688 #else
2689                 pkt_cnt++;
2690                 if ((indicate_max_pkts) && (pkt_cnt > indicate_max_pkts))
2691                         break;
2692 #endif
2693         }
2694         if (nic->lro) {
2695                 /* Clear all LRO sessions before exiting */
2696                 for (i=0; i<MAX_LRO_SESSIONS; i++) {
2697                         lro_t *lro = &nic->lro0_n[i];
2698                         if (lro->in_use) {
2699                                 update_L3L4_header(nic, lro);
2700                                 queue_rx_frame(lro->parent);
2701                                 clear_lro_session(lro);
2702                         }
2703                 }
2704         }
2705
2706         spin_unlock(&nic->rx_lock);
2707 }
2708
2709 /**
2710  *  tx_intr_handler - Transmit interrupt handler
2711  *  @nic : device private variable
2712  *  Description:
2713  *  If an interrupt was raised to indicate DMA complete of the
2714  *  Tx packet, this function is called. It identifies the last TxD
2715  *  whose buffer was freed and frees all skbs whose data have already
2716  *  DMA'ed into the NICs internal memory.
2717  *  Return Value:
2718  *  NONE
2719  */
2720
2721 static void tx_intr_handler(fifo_info_t *fifo_data)
2722 {
2723         nic_t *nic = fifo_data->nic;
2724         struct net_device *dev = (struct net_device *) nic->dev;
2725         tx_curr_get_info_t get_info, put_info;
2726         struct sk_buff *skb;
2727         TxD_t *txdlp;
2728
2729         get_info = fifo_data->tx_curr_get_info;
2730         put_info = fifo_data->tx_curr_put_info;
2731         txdlp = (TxD_t *) fifo_data->list_info[get_info.offset].
2732             list_virt_addr;
2733         while ((!(txdlp->Control_1 & TXD_LIST_OWN_XENA)) &&
2734                (get_info.offset != put_info.offset) &&
2735                (txdlp->Host_Control)) {
2736                 /* Check for TxD errors */
2737                 if (txdlp->Control_1 & TXD_T_CODE) {
2738                         unsigned long long err;
2739                         err = txdlp->Control_1 & TXD_T_CODE;
2740                         if ((err >> 48) == 0xA) {
2741                                 DBG_PRINT(TX_DBG, "TxD returned due \
2742 to loss of link\n");
2743                         }
2744                         else {
2745                                 DBG_PRINT(ERR_DBG, "***TxD error \
2746 %llx\n", err);
2747                         }
2748                 }
2749
2750                 skb = s2io_txdl_getskb(fifo_data, txdlp, get_info.offset);
2751                 if (skb == NULL) {
2752                         DBG_PRINT(ERR_DBG, "%s: Null skb ",
2753                         __FUNCTION__);
2754                         DBG_PRINT(ERR_DBG, "in Tx Free Intr\n");
2755                         return;
2756                 }
2757
2758                 /* Updating the statistics block */
2759                 nic->stats.tx_bytes += skb->len;
2760                 dev_kfree_skb_irq(skb);
2761
2762                 get_info.offset++;
2763                 get_info.offset %= get_info.fifo_len + 1;
2764                 txdlp = (TxD_t *) fifo_data->list_info
2765                     [get_info.offset].list_virt_addr;
2766                 fifo_data->tx_curr_get_info.offset =
2767                     get_info.offset;
2768         }
2769
2770         spin_lock(&nic->tx_lock);
2771         if (netif_queue_stopped(dev))
2772                 netif_wake_queue(dev);
2773         spin_unlock(&nic->tx_lock);
2774 }
2775
2776 /**
2777  *  alarm_intr_handler - Alarm Interrrupt handler
2778  *  @nic: device private variable
2779  *  Description: If the interrupt was neither because of Rx packet or Tx
2780  *  complete, this function is called. If the interrupt was to indicate
2781  *  a loss of link, the OSM link status handler is invoked for any other
2782  *  alarm interrupt the block that raised the interrupt is displayed
2783  *  and a H/W reset is issued.
2784  *  Return Value:
2785  *  NONE
2786 */
2787
2788 static void alarm_intr_handler(struct s2io_nic *nic)
2789 {
2790         struct net_device *dev = (struct net_device *) nic->dev;
2791         XENA_dev_config_t __iomem *bar0 = nic->bar0;
2792         register u64 val64 = 0, err_reg = 0;
2793
2794         /* Handling link status change error Intr */
2795         if (s2io_link_fault_indication(nic) == MAC_RMAC_ERR_TIMER) {
2796                 err_reg = readq(&bar0->mac_rmac_err_reg);
2797                 writeq(err_reg, &bar0->mac_rmac_err_reg);
2798                 if (err_reg & RMAC_LINK_STATE_CHANGE_INT) {
2799                         schedule_work(&nic->set_link_task);
2800                 }
2801         }
2802
2803         /* Handling Ecc errors */
2804         val64 = readq(&bar0->mc_err_reg);
2805         writeq(val64, &bar0->mc_err_reg);
2806         if (val64 & (MC_ERR_REG_ECC_ALL_SNG | MC_ERR_REG_ECC_ALL_DBL)) {
2807                 if (val64 & MC_ERR_REG_ECC_ALL_DBL) {
2808                         nic->mac_control.stats_info->sw_stat.
2809                                 double_ecc_errs++;
2810                         DBG_PRINT(INIT_DBG, "%s: Device indicates ",
2811                                   dev->name);
2812                         DBG_PRINT(INIT_DBG, "double ECC error!!\n");
2813                         if (nic->device_type != XFRAME_II_DEVICE) {
2814                                 /* Reset XframeI only if critical error */
2815                                 if (val64 & (MC_ERR_REG_MIRI_ECC_DB_ERR_0 |
2816                                              MC_ERR_REG_MIRI_ECC_DB_ERR_1)) {
2817                                         netif_stop_queue(dev);
2818                                         schedule_work(&nic->rst_timer_task);
2819                                 }
2820                         }
2821                 } else {
2822                         nic->mac_control.stats_info->sw_stat.
2823                                 single_ecc_errs++;
2824                 }
2825         }
2826
2827         /* In case of a serious error, the device will be Reset. */
2828         val64 = readq(&bar0->serr_source);
2829         if (val64 & SERR_SOURCE_ANY) {
2830                 DBG_PRINT(ERR_DBG, "%s: Device indicates ", dev->name);
2831                 DBG_PRINT(ERR_DBG, "serious error %llx!!\n",
2832                           (unsigned long long)val64);
2833                 netif_stop_queue(dev);
2834                 schedule_work(&nic->rst_timer_task);
2835         }
2836
2837         /*
2838          * Also as mentioned in the latest Errata sheets if the PCC_FB_ECC
2839          * Error occurs, the adapter will be recycled by disabling the
2840          * adapter enable bit and enabling it again after the device
2841          * becomes Quiescent.
2842          */
2843         val64 = readq(&bar0->pcc_err_reg);
2844         writeq(val64, &bar0->pcc_err_reg);
2845         if (val64 & PCC_FB_ECC_DB_ERR) {
2846                 u64 ac = readq(&bar0->adapter_control);
2847                 ac &= ~(ADAPTER_CNTL_EN);
2848                 writeq(ac, &bar0->adapter_control);
2849                 ac = readq(&bar0->adapter_control);
2850                 schedule_work(&nic->set_link_task);
2851         }
2852
2853         /* Other type of interrupts are not being handled now,  TODO */
2854 }
2855
2856 /**
2857  *  wait_for_cmd_complete - waits for a command to complete.
2858  *  @sp : private member of the device structure, which is a pointer to the
2859  *  s2io_nic structure.
2860  *  Description: Function that waits for a command to Write into RMAC
2861  *  ADDR DATA registers to be completed and returns either success or
2862  *  error depending on whether the command was complete or not.
2863  *  Return value:
2864  *   SUCCESS on success and FAILURE on failure.
2865  */
2866
2867 static int wait_for_cmd_complete(nic_t * sp)
2868 {
2869         XENA_dev_config_t __iomem *bar0 = sp->bar0;
2870         int ret = FAILURE, cnt = 0;
2871         u64 val64;
2872
2873         while (TRUE) {
2874                 val64 = readq(&bar0->rmac_addr_cmd_mem);
2875                 if (!(val64 & RMAC_ADDR_CMD_MEM_STROBE_CMD_EXECUTING)) {
2876                         ret = SUCCESS;
2877                         break;
2878                 }
2879                 msleep(50);
2880                 if (cnt++ > 10)
2881                         break;
2882         }
2883
2884         return ret;
2885 }
2886
2887 /**
2888  *  s2io_reset - Resets the card.
2889  *  @sp : private member of the device structure.
2890  *  Description: Function to Reset the card. This function then also
2891  *  restores the previously saved PCI configuration space registers as
2892  *  the card reset also resets the configuration space.
2893  *  Return value:
2894  *  void.
2895  */
2896
2897 static void s2io_reset(nic_t * sp)
2898 {
2899         XENA_dev_config_t __iomem *bar0 = sp->bar0;
2900         u64 val64;
2901         u16 subid, pci_cmd;
2902
2903         /* Back up  the PCI-X CMD reg, dont want to lose MMRBC, OST settings */
2904         pci_read_config_word(sp->pdev, PCIX_COMMAND_REGISTER, &(pci_cmd));
2905
2906         val64 = SW_RESET_ALL;
2907         writeq(val64, &bar0->sw_reset);
2908
2909         /*
2910          * At this stage, if the PCI write is indeed completed, the
2911          * card is reset and so is the PCI Config space of the device.
2912          * So a read cannot be issued at this stage on any of the
2913          * registers to ensure the write into "sw_reset" register
2914          * has gone through.
2915          * Question: Is there any system call that will explicitly force
2916          * all the write commands still pending on the bus to be pushed
2917          * through?
2918          * As of now I'am just giving a 250ms delay and hoping that the
2919          * PCI write to sw_reset register is done by this time.
2920          */
2921         msleep(250);
2922
2923         /* Restore the PCI state saved during initialization. */
2924         pci_restore_state(sp->pdev);
2925         pci_write_config_word(sp->pdev, PCIX_COMMAND_REGISTER,
2926                                      pci_cmd);
2927         s2io_init_pci(sp);
2928
2929         msleep(250);
2930
2931         /* Set swapper to enable I/O register access */
2932         s2io_set_swapper(sp);
2933
2934         /* Restore the MSIX table entries from local variables */
2935         restore_xmsi_data(sp);
2936
2937         /* Clear certain PCI/PCI-X fields after reset */
2938         if (sp->device_type == XFRAME_II_DEVICE) {
2939                 /* Clear parity err detect bit */
2940                 pci_write_config_word(sp->pdev, PCI_STATUS, 0x8000);
2941
2942                 /* Clearing PCIX Ecc status register */
2943                 pci_write_config_dword(sp->pdev, 0x68, 0x7C);
2944
2945                 /* Clearing PCI_STATUS error reflected here */
2946                 writeq(BIT(62), &bar0->txpic_int_reg);
2947         }
2948
2949         /* Reset device statistics maintained by OS */
2950         memset(&sp->stats, 0, sizeof (struct net_device_stats));
2951
2952         /* SXE-002: Configure link and activity LED to turn it off */
2953         subid = sp->pdev->subsystem_device;
2954         if (((subid & 0xFF) >= 0x07) &&
2955             (sp->device_type == XFRAME_I_DEVICE)) {
2956                 val64 = readq(&bar0->gpio_control);
2957                 val64 |= 0x0000800000000000ULL;
2958                 writeq(val64, &bar0->gpio_control);
2959                 val64 = 0x0411040400000000ULL;
2960                 writeq(val64, (void __iomem *)bar0 + 0x2700);
2961         }
2962
2963         /*
2964          * Clear spurious ECC interrupts that would have occured on
2965          * XFRAME II cards after reset.
2966          */
2967         if (sp->device_type == XFRAME_II_DEVICE) {
2968                 val64 = readq(&bar0->pcc_err_reg);
2969                 writeq(val64, &bar0->pcc_err_reg);
2970         }
2971
2972         sp->device_enabled_once = FALSE;
2973 }
2974
2975 /**
2976  *  s2io_set_swapper - to set the swapper controle on the card
2977  *  @sp : private member of the device structure,
2978  *  pointer to the s2io_nic structure.
2979  *  Description: Function to set the swapper control on the card
2980  *  correctly depending on the 'endianness' of the system.
2981  *  Return value:
2982  *  SUCCESS on success and FAILURE on failure.
2983  */
2984
2985 static int s2io_set_swapper(nic_t * sp)
2986 {
2987         struct net_device *dev = sp->dev;
2988         XENA_dev_config_t __iomem *bar0 = sp->bar0;
2989         u64 val64, valt, valr;
2990
2991         /*
2992          * Set proper endian settings and verify the same by reading
2993          * the PIF Feed-back register.
2994          */
2995
2996         val64 = readq(&bar0->pif_rd_swapper_fb);
2997         if (val64 != 0x0123456789ABCDEFULL) {
2998                 int i = 0;
2999                 u64 value[] = { 0xC30000C3C30000C3ULL,   /* FE=1, SE=1 */
3000                                 0x8100008181000081ULL,  /* FE=1, SE=0 */
3001                                 0x4200004242000042ULL,  /* FE=0, SE=1 */
3002                                 0};                     /* FE=0, SE=0 */
3003
3004                 while(i<4) {
3005                         writeq(value[i], &bar0->swapper_ctrl);
3006                         val64 = readq(&bar0->pif_rd_swapper_fb);
3007                         if (val64 == 0x0123456789ABCDEFULL)
3008                                 break;
3009                         i++;
3010                 }
3011                 if (i == 4) {
3012                         DBG_PRINT(ERR_DBG, "%s: Endian settings are wrong, ",
3013                                 dev->name);
3014                         DBG_PRINT(ERR_DBG, "feedback read %llx\n",
3015                                 (unsigned long long) val64);
3016                         return FAILURE;
3017                 }
3018                 valr = value[i];
3019         } else {
3020                 valr = readq(&bar0->swapper_ctrl);
3021         }
3022
3023         valt = 0x0123456789ABCDEFULL;
3024         writeq(valt, &bar0->xmsi_address);
3025         val64 = readq(&bar0->xmsi_address);
3026
3027         if(val64 != valt) {
3028                 int i = 0;
3029                 u64 value[] = { 0x00C3C30000C3C300ULL,  /* FE=1, SE=1 */
3030                                 0x0081810000818100ULL,  /* FE=1, SE=0 */
3031                                 0x0042420000424200ULL,  /* FE=0, SE=1 */
3032                                 0};                     /* FE=0, SE=0 */
3033
3034                 while(i<4) {
3035                         writeq((value[i] | valr), &bar0->swapper_ctrl);
3036                         writeq(valt, &bar0->xmsi_address);
3037                         val64 = readq(&bar0->xmsi_address);
3038                         if(val64 == valt)
3039                                 break;
3040                         i++;
3041                 }
3042                 if(i == 4) {
3043                         unsigned long long x = val64;
3044                         DBG_PRINT(ERR_DBG, "Write failed, Xmsi_addr ");
3045                         DBG_PRINT(ERR_DBG, "reads:0x%llx\n", x);
3046                         return FAILURE;
3047                 }
3048         }
3049         val64 = readq(&bar0->swapper_ctrl);
3050         val64 &= 0xFFFF000000000000ULL;
3051
3052 #ifdef  __BIG_ENDIAN
3053         /*
3054          * The device by default set to a big endian format, so a
3055          * big endian driver need not set anything.
3056          */
3057         val64 |= (SWAPPER_CTRL_TXP_FE |
3058                  SWAPPER_CTRL_TXP_SE |
3059                  SWAPPER_CTRL_TXD_R_FE |
3060                  SWAPPER_CTRL_TXD_W_FE |
3061                  SWAPPER_CTRL_TXF_R_FE |
3062                  SWAPPER_CTRL_RXD_R_FE |
3063                  SWAPPER_CTRL_RXD_W_FE |
3064                  SWAPPER_CTRL_RXF_W_FE |
3065                  SWAPPER_CTRL_XMSI_FE |
3066                  SWAPPER_CTRL_STATS_FE | SWAPPER_CTRL_STATS_SE);
3067         if (sp->intr_type == INTA)
3068                 val64 |= SWAPPER_CTRL_XMSI_SE;
3069         writeq(val64, &bar0->swapper_ctrl);
3070 #else
3071         /*
3072          * Initially we enable all bits to make it accessible by the
3073          * driver, then we selectively enable only those bits that
3074          * we want to set.
3075          */
3076         val64 |= (SWAPPER_CTRL_TXP_FE |
3077                  SWAPPER_CTRL_TXP_SE |
3078                  SWAPPER_CTRL_TXD_R_FE |
3079                  SWAPPER_CTRL_TXD_R_SE |
3080                  SWAPPER_CTRL_TXD_W_FE |
3081                  SWAPPER_CTRL_TXD_W_SE |
3082                  SWAPPER_CTRL_TXF_R_FE |
3083                  SWAPPER_CTRL_RXD_R_FE |
3084                  SWAPPER_CTRL_RXD_R_SE |
3085                  SWAPPER_CTRL_RXD_W_FE |
3086                  SWAPPER_CTRL_RXD_W_SE |
3087                  SWAPPER_CTRL_RXF_W_FE |
3088                  SWAPPER_CTRL_XMSI_FE |
3089                  SWAPPER_CTRL_STATS_FE | SWAPPER_CTRL_STATS_SE);
3090         if (sp->intr_type == INTA)
3091                 val64 |= SWAPPER_CTRL_XMSI_SE;
3092         writeq(val64, &bar0->swapper_ctrl);
3093 #endif
3094         val64 = readq(&bar0->swapper_ctrl);
3095
3096         /*
3097          * Verifying if endian settings are accurate by reading a
3098          * feedback register.
3099          */
3100         val64 = readq(&bar0->pif_rd_swapper_fb);
3101         if (val64 != 0x0123456789ABCDEFULL) {
3102                 /* Endian settings are incorrect, calls for another dekko. */
3103                 DBG_PRINT(ERR_DBG, "%s: Endian settings are wrong, ",
3104                           dev->name);
3105                 DBG_PRINT(ERR_DBG, "feedback read %llx\n",
3106                           (unsigned long long) val64);
3107                 return FAILURE;
3108         }
3109
3110         return SUCCESS;
3111 }
3112
3113 static int wait_for_msix_trans(nic_t *nic, int i)
3114 {
3115         XENA_dev_config_t __iomem *bar0 = nic->bar0;
3116         u64 val64;
3117         int ret = 0, cnt = 0;
3118
3119         do {
3120                 val64 = readq(&bar0->xmsi_access);
3121                 if (!(val64 & BIT(15)))
3122                         break;
3123                 mdelay(1);
3124                 cnt++;
3125         } while(cnt < 5);
3126         if (cnt == 5) {
3127                 DBG_PRINT(ERR_DBG, "XMSI # %d Access failed\n", i);
3128                 ret = 1;
3129         }
3130
3131         return ret;
3132 }
3133
3134 static void restore_xmsi_data(nic_t *nic)
3135 {
3136         XENA_dev_config_t __iomem *bar0 = nic->bar0;
3137         u64 val64;
3138         int i;
3139
3140         for (i=0; i< MAX_REQUESTED_MSI_X; i++) {
3141                 writeq(nic->msix_info[i].addr, &bar0->xmsi_address);
3142                 writeq(nic->msix_info[i].data, &bar0->xmsi_data);
3143                 val64 = (BIT(7) | BIT(15) | vBIT(i, 26, 6));
3144                 writeq(val64, &bar0->xmsi_access);
3145                 if (wait_for_msix_trans(nic, i)) {
3146                         DBG_PRINT(ERR_DBG, "failed in %s\n", __FUNCTION__);
3147                         continue;
3148                 }
3149         }
3150 }
3151
3152 static void store_xmsi_data(nic_t *nic)
3153 {
3154         XENA_dev_config_t __iomem *bar0 = nic->bar0;
3155         u64 val64, addr, data;
3156         int i;
3157
3158         /* Store and display */
3159         for (i=0; i< MAX_REQUESTED_MSI_X; i++) {
3160                 val64 = (BIT(15) | vBIT(i, 26, 6));
3161                 writeq(val64, &bar0->xmsi_access);
3162                 if (wait_for_msix_trans(nic, i)) {
3163                         DBG_PRINT(ERR_DBG, "failed in %s\n", __FUNCTION__);
3164                         continue;
3165                 }
3166                 addr = readq(&bar0->xmsi_address);
3167                 data = readq(&bar0->xmsi_data);
3168                 if (addr && data) {
3169                         nic->msix_info[i].addr = addr;
3170                         nic->msix_info[i].data = data;
3171                 }
3172         }
3173 }
3174
3175 int s2io_enable_msi(nic_t *nic)
3176 {
3177         XENA_dev_config_t __iomem *bar0 = nic->bar0;
3178         u16 msi_ctrl, msg_val;
3179         struct config_param *config = &nic->config;
3180         struct net_device *dev = nic->dev;
3181         u64 val64, tx_mat, rx_mat;
3182         int i, err;
3183
3184         val64 = readq(&bar0->pic_control);
3185         val64 &= ~BIT(1);
3186         writeq(val64, &bar0->pic_control);
3187
3188         err = pci_enable_msi(nic->pdev);
3189         if (err) {
3190                 DBG_PRINT(ERR_DBG, "%s: enabling MSI failed\n",
3191                           nic->dev->name);
3192                 return err;
3193         }
3194
3195         /*
3196          * Enable MSI and use MSI-1 in stead of the standard MSI-0
3197          * for interrupt handling.
3198          */
3199         pci_read_config_word(nic->pdev, 0x4c, &msg_val);
3200         msg_val ^= 0x1;
3201         pci_write_config_word(nic->pdev, 0x4c, msg_val);
3202         pci_read_config_word(nic->pdev, 0x4c, &msg_val);
3203
3204         pci_read_config_word(nic->pdev, 0x42, &msi_ctrl);
3205         msi_ctrl |= 0x10;
3206         pci_write_config_word(nic->pdev, 0x42, msi_ctrl);
3207
3208         /* program MSI-1 into all usable Tx_Mat and Rx_Mat fields */
3209         tx_mat = readq(&bar0->tx_mat0_n[0]);
3210         for (i=0; i<config->tx_fifo_num; i++) {
3211                 tx_mat |= TX_MAT_SET(i, 1);
3212         }
3213         writeq(tx_mat, &bar0->tx_mat0_n[0]);
3214
3215         rx_mat = readq(&bar0->rx_mat);
3216         for (i=0; i<config->rx_ring_num; i++) {
3217                 rx_mat |= RX_MAT_SET(i, 1);
3218         }
3219         writeq(rx_mat, &bar0->rx_mat);
3220
3221         dev->irq = nic->pdev->irq;
3222         return 0;
3223 }
3224
3225 static int s2io_enable_msi_x(nic_t *nic)
3226 {
3227         XENA_dev_config_t __iomem *bar0 = nic->bar0;
3228         u64 tx_mat, rx_mat;
3229         u16 msi_control; /* Temp variable */
3230         int ret, i, j, msix_indx = 1;
3231
3232         nic->entries = kmalloc(MAX_REQUESTED_MSI_X * sizeof(struct msix_entry),
3233                                GFP_KERNEL);
3234         if (nic->entries == NULL) {
3235                 DBG_PRINT(ERR_DBG, "%s: Memory allocation failed\n", __FUNCTION__);
3236                 return -ENOMEM;
3237         }
3238         memset(nic->entries, 0, MAX_REQUESTED_MSI_X * sizeof(struct msix_entry));
3239
3240         nic->s2io_entries =
3241                 kmalloc(MAX_REQUESTED_MSI_X * sizeof(struct s2io_msix_entry),
3242                                    GFP_KERNEL);
3243         if (nic->s2io_entries == NULL) {
3244                 DBG_PRINT(ERR_DBG, "%s: Memory allocation failed\n", __FUNCTION__);
3245                 kfree(nic->entries);
3246                 return -ENOMEM;
3247         }
3248         memset(nic->s2io_entries, 0,
3249                MAX_REQUESTED_MSI_X * sizeof(struct s2io_msix_entry));
3250
3251         for (i=0; i< MAX_REQUESTED_MSI_X; i++) {
3252                 nic->entries[i].entry = i;
3253                 nic->s2io_entries[i].entry = i;
3254                 nic->s2io_entries[i].arg = NULL;
3255                 nic->s2io_entries[i].in_use = 0;
3256         }
3257
3258         tx_mat = readq(&bar0->tx_mat0_n[0]);
3259         for (i=0; i<nic->config.tx_fifo_num; i++, msix_indx++) {
3260                 tx_mat |= TX_MAT_SET(i, msix_indx);
3261                 nic->s2io_entries[msix_indx].arg = &nic->mac_control.fifos[i];
3262                 nic->s2io_entries[msix_indx].type = MSIX_FIFO_TYPE;
3263                 nic->s2io_entries[msix_indx].in_use = MSIX_FLG;
3264         }
3265         writeq(tx_mat, &bar0->tx_mat0_n[0]);
3266
3267         if (!nic->config.bimodal) {
3268                 rx_mat = readq(&bar0->rx_mat);
3269                 for (j=0; j<nic->config.rx_ring_num; j++, msix_indx++) {
3270                         rx_mat |= RX_MAT_SET(j, msix_indx);
3271                         nic->s2io_entries[msix_indx].arg = &nic->mac_control.rings[j];
3272                         nic->s2io_entries[msix_indx].type = MSIX_RING_TYPE;
3273                         nic->s2io_entries[msix_indx].in_use = MSIX_FLG;
3274                 }
3275                 writeq(rx_mat, &bar0->rx_mat);
3276         } else {
3277                 tx_mat = readq(&bar0->tx_mat0_n[7]);
3278                 for (j=0; j<nic->config.rx_ring_num; j++, msix_indx++) {
3279                         tx_mat |= TX_MAT_SET(i, msix_indx);
3280                         nic->s2io_entries[msix_indx].arg = &nic->mac_control.rings[j];
3281                         nic->s2io_entries[msix_indx].type = MSIX_RING_TYPE;
3282                         nic->s2io_entries[msix_indx].in_use = MSIX_FLG;
3283                 }
3284                 writeq(tx_mat, &bar0->tx_mat0_n[7]);
3285         }
3286
3287         ret = pci_enable_msix(nic->pdev, nic->entries, MAX_REQUESTED_MSI_X);
3288         if (ret) {
3289                 DBG_PRINT(ERR_DBG, "%s: Enabling MSIX failed\n", nic->dev->name);
3290                 kfree(nic->entries);
3291                 kfree(nic->s2io_entries);
3292                 nic->entries = NULL;
3293                 nic->s2io_entries = NULL;
3294                 return -ENOMEM;
3295         }
3296
3297         /*
3298          * To enable MSI-X, MSI also needs to be enabled, due to a bug
3299          * in the herc NIC. (Temp change, needs to be removed later)
3300          */
3301         pci_read_config_word(nic->pdev, 0x42, &msi_control);
3302         msi_control |= 0x1; /* Enable MSI */
3303         pci_write_config_word(nic->pdev, 0x42, msi_control);
3304
3305         return 0;
3306 }
3307
3308 /* ********************************************************* *
3309  * Functions defined below concern the OS part of the driver *
3310  * ********************************************************* */
3311
3312 /**
3313  *  s2io_open - open entry point of the driver
3314  *  @dev : pointer to the device structure.
3315  *  Description:
3316  *  This function is the open entry point of the driver. It mainly calls a
3317  *  function to allocate Rx buffers and inserts them into the buffer
3318  *  descriptors and then enables the Rx part of the NIC.
3319  *  Return value:
3320  *  0 on success and an appropriate (-)ve integer as defined in errno.h
3321  *   file on failure.
3322  */
3323
3324 static int s2io_open(struct net_device *dev)
3325 {
3326         nic_t *sp = dev->priv;
3327         int err = 0;
3328         int i;
3329         u16 msi_control; /* Temp variable */
3330
3331         /*
3332          * Make sure you have link off by default every time
3333          * Nic is initialized
3334          */
3335         netif_carrier_off(dev);
3336         sp->last_link_state = 0;
3337
3338         /* Initialize H/W and enable interrupts */
3339         if (s2io_card_up(sp)) {
3340                 DBG_PRINT(ERR_DBG, "%s: H/W initialization failed\n",
3341                           dev->name);
3342                 err = -ENODEV;
3343                 goto hw_init_failed;
3344         }
3345
3346         /* Store the values of the MSIX table in the nic_t structure */
3347         store_xmsi_data(sp);
3348
3349         /* After proper initialization of H/W, register ISR */
3350         if (sp->intr_type == MSI) {
3351                 err = request_irq((int) sp->pdev->irq, s2io_msi_handle,
3352                         SA_SHIRQ, sp->name, dev);
3353                 if (err) {
3354                         DBG_PRINT(ERR_DBG, "%s: MSI registration \
3355 failed\n", dev->name);
3356                         goto isr_registration_failed;
3357                 }
3358         }
3359         if (sp->intr_type == MSI_X) {
3360                 for (i=1; (sp->s2io_entries[i].in_use == MSIX_FLG); i++) {
3361                         if (sp->s2io_entries[i].type == MSIX_FIFO_TYPE) {
3362                                 sprintf(sp->desc1, "%s:MSI-X-%d-TX",
3363                                         dev->name, i);
3364                                 err = request_irq(sp->entries[i].vector,
3365                                           s2io_msix_fifo_handle, 0, sp->desc1,
3366                                           sp->s2io_entries[i].arg);
3367                                 DBG_PRINT(ERR_DBG, "%s @ 0x%llx\n", sp->desc1,
3368                                     (unsigned long long)sp->msix_info[i].addr);
3369                         } else {
3370                                 sprintf(sp->desc2, "%s:MSI-X-%d-RX",
3371                                         dev->name, i);
3372                                 err = request_irq(sp->entries[i].vector,
3373                                           s2io_msix_ring_handle, 0, sp->desc2,
3374                                           sp->s2io_entries[i].arg);
3375                                 DBG_PRINT(ERR_DBG, "%s @ 0x%llx\n", sp->desc2,
3376                                      (unsigned long long)sp->msix_info[i].addr);
3377                         }
3378                         if (err) {
3379                                 DBG_PRINT(ERR_DBG, "%s: MSI-X-%d registration \
3380 failed\n", dev->name, i);
3381                                 DBG_PRINT(ERR_DBG, "Returned: %d\n", err);
3382                                 goto isr_registration_failed;
3383                         }
3384                         sp->s2io_entries[i].in_use = MSIX_REGISTERED_SUCCESS;
3385                 }
3386         }
3387         if (sp->intr_type == INTA) {
3388                 err = request_irq((int) sp->pdev->irq, s2io_isr, SA_SHIRQ,
3389                                 sp->name, dev);
3390                 if (err) {
3391                         DBG_PRINT(ERR_DBG, "%s: ISR registration failed\n",
3392                                   dev->name);
3393                         goto isr_registration_failed;
3394                 }
3395         }
3396
3397         if (s2io_set_mac_addr(dev, dev->dev_addr) == FAILURE) {
3398                 DBG_PRINT(ERR_DBG, "Set Mac Address Failed\n");
3399                 err = -ENODEV;
3400                 goto setting_mac_address_failed;
3401         }
3402
3403         netif_start_queue(dev);
3404         return 0;
3405
3406 setting_mac_address_failed:
3407         if (sp->intr_type != MSI_X)
3408                 free_irq(sp->pdev->irq, dev);
3409 isr_registration_failed:
3410         del_timer_sync(&sp->alarm_timer);
3411         if (sp->intr_type == MSI_X) {
3412                 if (sp->device_type == XFRAME_II_DEVICE) {
3413                         for (i=1; (sp->s2io_entries[i].in_use ==
3414                                 MSIX_REGISTERED_SUCCESS); i++) {
3415                                 int vector = sp->entries[i].vector;
3416                                 void *arg = sp->s2io_entries[i].arg;
3417
3418                                 free_irq(vector, arg);
3419                         }
3420                         pci_disable_msix(sp->pdev);
3421
3422                         /* Temp */
3423                         pci_read_config_word(sp->pdev, 0x42, &msi_control);
3424                         msi_control &= 0xFFFE; /* Disable MSI */
3425                         pci_write_config_word(sp->pdev, 0x42, msi_control);
3426                 }
3427         }
3428         else if (sp->intr_type == MSI)
3429                 pci_disable_msi(sp->pdev);
3430         s2io_reset(sp);
3431 hw_init_failed:
3432         if (sp->intr_type == MSI_X) {
3433                 if (sp->entries)
3434                         kfree(sp->entries);
3435                 if (sp->s2io_entries)
3436                         kfree(sp->s2io_entries);
3437         }
3438         return err;
3439 }
3440
3441 /**
3442  *  s2io_close -close entry point of the driver
3443  *  @dev : device pointer.
3444  *  Description:
3445  *  This is the stop entry point of the driver. It needs to undo exactly
3446  *  whatever was done by the open entry point,thus it's usually referred to
3447  *  as the close function.Among other things this function mainly stops the
3448  *  Rx side of the NIC and frees all the Rx buffers in the Rx rings.
3449  *  Return value:
3450  *  0 on success and an appropriate (-)ve integer as defined in errno.h
3451  *  file on failure.
3452  */
3453
3454 static int s2io_close(struct net_device *dev)
3455 {
3456         nic_t *sp = dev->priv;
3457         int i;
3458         u16 msi_control;
3459
3460         flush_scheduled_work();
3461         netif_stop_queue(dev);
3462         /* Reset card, kill tasklet and free Tx and Rx buffers. */
3463         s2io_card_down(sp);
3464
3465         if (sp->intr_type == MSI_X) {
3466                 if (sp->device_type == XFRAME_II_DEVICE) {
3467                         for (i=1; (sp->s2io_entries[i].in_use ==
3468                                         MSIX_REGISTERED_SUCCESS); i++) {
3469                                 int vector = sp->entries[i].vector;
3470                                 void *arg = sp->s2io_entries[i].arg;
3471
3472                                 free_irq(vector, arg);
3473                         }
3474                         pci_read_config_word(sp->pdev, 0x42, &msi_control);
3475                         msi_control &= 0xFFFE; /* Disable MSI */
3476                         pci_write_config_word(sp->pdev, 0x42, msi_control);
3477
3478                         pci_disable_msix(sp->pdev);
3479                 }
3480         }
3481         else {
3482                 free_irq(sp->pdev->irq, dev);
3483                 if (sp->intr_type == MSI)
3484                         pci_disable_msi(sp->pdev);
3485         }
3486         sp->device_close_flag = TRUE;   /* Device is shut down. */
3487         return 0;
3488 }
3489
3490 /**
3491  *  s2io_xmit - Tx entry point of te driver
3492  *  @skb : the socket buffer containing the Tx data.
3493  *  @dev : device pointer.
3494  *  Description :
3495  *  This function is the Tx entry point of the driver. S2IO NIC supports
3496  *  certain protocol assist features on Tx side, namely  CSO, S/G, LSO.
3497  *  NOTE: when device cant queue the pkt,just the trans_start variable will
3498  *  not be upadted.
3499  *  Return value:
3500  *  0 on success & 1 on failure.
3501  */
3502
3503 static int s2io_xmit(struct sk_buff *skb, struct net_device *dev)
3504 {
3505         nic_t *sp = dev->priv;
3506         u16 frg_cnt, frg_len, i, queue, queue_len, put_off, get_off;
3507         register u64 val64;
3508         TxD_t *txdp;
3509         TxFIFO_element_t __iomem *tx_fifo;
3510         unsigned long flags;
3511 #ifdef NETIF_F_TSO
3512         int mss;
3513 #endif
3514         u16 vlan_tag = 0;
3515         int vlan_priority = 0;
3516         mac_info_t *mac_control;
3517         struct config_param *config;
3518
3519         mac_control = &sp->mac_control;
3520         config = &sp->config;
3521
3522         DBG_PRINT(TX_DBG, "%s: In Neterion Tx routine\n", dev->name);
3523         spin_lock_irqsave(&sp->tx_lock, flags);
3524         if (atomic_read(&sp->card_state) == CARD_DOWN) {
3525                 DBG_PRINT(TX_DBG, "%s: Card going down for reset\n",
3526                           dev->name);
3527                 spin_unlock_irqrestore(&sp->tx_lock, flags);
3528                 dev_kfree_skb(skb);
3529                 return 0;
3530         }
3531
3532         queue = 0;
3533
3534         /* Get Fifo number to Transmit based on vlan priority */
3535         if (sp->vlgrp && vlan_tx_tag_present(skb)) {
3536                 vlan_tag = vlan_tx_tag_get(skb);
3537                 vlan_priority = vlan_tag >> 13;
3538                 queue = config->fifo_mapping[vlan_priority];
3539         }
3540
3541         put_off = (u16) mac_control->fifos[queue].tx_curr_put_info.offset;
3542         get_off = (u16) mac_control->fifos[queue].tx_curr_get_info.offset;
3543         txdp = (TxD_t *) mac_control->fifos[queue].list_info[put_off].
3544                 list_virt_addr;
3545
3546         queue_len = mac_control->fifos[queue].tx_curr_put_info.fifo_len + 1;
3547         /* Avoid "put" pointer going beyond "get" pointer */
3548         if (txdp->Host_Control || (((put_off + 1) % queue_len) == get_off)) {
3549                 DBG_PRINT(TX_DBG, "Error in xmit, No free TXDs.\n");
3550                 netif_stop_queue(dev);
3551                 dev_kfree_skb(skb);
3552                 spin_unlock_irqrestore(&sp->tx_lock, flags);
3553                 return 0;
3554         }
3555
3556         /* A buffer with no data will be dropped */
3557         if (!skb->len) {
3558                 DBG_PRINT(TX_DBG, "%s:Buffer has no data..\n", dev->name);
3559                 dev_kfree_skb(skb);
3560                 spin_unlock_irqrestore(&sp->tx_lock, flags);
3561                 return 0;
3562         }
3563
3564         txdp->Control_1 = 0;
3565         txdp->Control_2 = 0;
3566 #ifdef NETIF_F_TSO
3567         mss = skb_shinfo(skb)->tso_size;
3568         if (mss) {
3569                 txdp->Control_1 |= TXD_TCP_LSO_EN;
3570                 txdp->Control_1 |= TXD_TCP_LSO_MSS(mss);
3571         }
3572 #endif
3573         if (skb->ip_summed == CHECKSUM_HW) {
3574                 txdp->Control_2 |=
3575                     (TXD_TX_CKO_IPV4_EN | TXD_TX_CKO_TCP_EN |
3576                      TXD_TX_CKO_UDP_EN);
3577         }
3578         txdp->Control_1 |= TXD_GATHER_CODE_FIRST;
3579         txdp->Control_1 |= TXD_LIST_OWN_XENA;
3580         txdp->Control_2 |= config->tx_intr_type;
3581
3582         if (sp->vlgrp && vlan_tx_tag_present(skb)) {
3583                 txdp->Control_2 |= TXD_VLAN_ENABLE;
3584                 txdp->Control_2 |= TXD_VLAN_TAG(vlan_tag);
3585         }
3586
3587         frg_len = skb->len - skb->data_len;
3588         if (skb_shinfo(skb)->ufo_size) {
3589                 int ufo_size;
3590
3591                 ufo_size = skb_shinfo(skb)->ufo_size;
3592                 ufo_size &= ~7;
3593                 txdp->Control_1 |= TXD_UFO_EN;
3594                 txdp->Control_1 |= TXD_UFO_MSS(ufo_size);
3595                 txdp->Control_1 |= TXD_BUFFER0_SIZE(8);
3596 #ifdef __BIG_ENDIAN
3597                 sp->ufo_in_band_v[put_off] =
3598                                 (u64)skb_shinfo(skb)->ip6_frag_id;
3599 #else
3600                 sp->ufo_in_band_v[put_off] =
3601                                 (u64)skb_shinfo(skb)->ip6_frag_id << 32;
3602 #endif
3603                 txdp->Host_Control = (unsigned long)sp->ufo_in_band_v;
3604                 txdp->Buffer_Pointer = pci_map_single(sp->pdev,
3605                                         sp->ufo_in_band_v,
3606                                         sizeof(u64), PCI_DMA_TODEVICE);
3607                 txdp++;
3608                 txdp->Control_1 = 0;
3609                 txdp->Control_2 = 0;
3610         }
3611
3612         txdp->Buffer_Pointer = pci_map_single
3613             (sp->pdev, skb->data, frg_len, PCI_DMA_TODEVICE);
3614         txdp->Host_Control = (unsigned long) skb;
3615         txdp->Control_1 |= TXD_BUFFER0_SIZE(frg_len);
3616
3617         if (skb_shinfo(skb)->ufo_size)
3618                 txdp->Control_1 |= TXD_UFO_EN;
3619
3620         frg_cnt = skb_shinfo(skb)->nr_frags;
3621         /* For fragmented SKB. */
3622         for (i = 0; i < frg_cnt; i++) {
3623                 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
3624                 /* A '0' length fragment will be ignored */
3625                 if (!frag->size)
3626                         continue;
3627                 txdp++;
3628                 txdp->Buffer_Pointer = (u64) pci_map_page
3629                     (sp->pdev, frag->page, frag->page_offset,
3630                      frag->size, PCI_DMA_TODEVICE);
3631                 txdp->Control_1 = TXD_BUFFER0_SIZE(frag->size);
3632                 if (skb_shinfo(skb)->ufo_size)
3633                         txdp->Control_1 |= TXD_UFO_EN;
3634         }
3635         txdp->Control_1 |= TXD_GATHER_CODE_LAST;
3636
3637         if (skb_shinfo(skb)->ufo_size)
3638                 frg_cnt++; /* as Txd0 was used for inband header */
3639
3640         tx_fifo = mac_control->tx_FIFO_start[queue];
3641         val64 = mac_control->fifos[queue].list_info[put_off].list_phy_addr;
3642         writeq(val64, &tx_fifo->TxDL_Pointer);
3643
3644         val64 = (TX_FIFO_LAST_TXD_NUM(frg_cnt) | TX_FIFO_FIRST_LIST |
3645                  TX_FIFO_LAST_LIST);
3646
3647 #ifdef NETIF_F_TSO
3648         if (mss)
3649                 val64 |= TX_FIFO_SPECIAL_FUNC;
3650 #endif
3651         if (skb_shinfo(skb)->ufo_size)
3652                 val64 |= TX_FIFO_SPECIAL_FUNC;
3653         writeq(val64, &tx_fifo->List_Control);
3654
3655         mmiowb();
3656
3657         put_off++;
3658         put_off %= mac_control->fifos[queue].tx_curr_put_info.fifo_len + 1;
3659         mac_control->fifos[queue].tx_curr_put_info.offset = put_off;
3660
3661         /* Avoid "put" pointer going beyond "get" pointer */
3662         if (((put_off + 1) % queue_len) == get_off) {
3663                 DBG_PRINT(TX_DBG,
3664                           "No free TxDs for xmit, Put: 0x%x Get:0x%x\n",
3665                           put_off, get_off);
3666                 netif_stop_queue(dev);
3667         }
3668
3669         dev->trans_start = jiffies;
3670         spin_unlock_irqrestore(&sp->tx_lock, flags);
3671
3672         return 0;
3673 }
3674
3675 static void
3676 s2io_alarm_handle(unsigned long data)
3677 {
3678         nic_t *sp = (nic_t *)data;
3679
3680         alarm_intr_handler(sp);
3681         mod_timer(&sp->alarm_timer, jiffies + HZ / 2);
3682 }
3683
3684 static irqreturn_t
3685 s2io_msi_handle(int irq, void *dev_id, struct pt_regs *regs)
3686 {
3687         struct net_device *dev = (struct net_device *) dev_id;
3688         nic_t *sp = dev->priv;
3689         int i;
3690         int ret;
3691         mac_info_t *mac_control;
3692         struct config_param *config;
3693
3694         atomic_inc(&sp->isr_cnt);
3695         mac_control = &sp->mac_control;
3696         config = &sp->config;
3697         DBG_PRINT(INTR_DBG, "%s: MSI handler\n", __FUNCTION__);
3698
3699         /* If Intr is because of Rx Traffic */
3700         for (i = 0; i < config->rx_ring_num; i++)
3701                 rx_intr_handler(&mac_control->rings[i]);
3702
3703         /* If Intr is because of Tx Traffic */
3704         for (i = 0; i < config->tx_fifo_num; i++)
3705                 tx_intr_handler(&mac_control->fifos[i]);
3706
3707         /*
3708          * If the Rx buffer count is below the panic threshold then
3709          * reallocate the buffers from the interrupt handler itself,
3710          * else schedule a tasklet to reallocate the buffers.
3711          */
3712         for (i = 0; i < config->rx_ring_num; i++) {
3713                 if (!sp->lro) {
3714                         int rxb_size = atomic_read(&sp->rx_bufs_left[i]);
3715                         int level = rx_buffer_level(sp, rxb_size, i);
3716
3717                         if ((level == PANIC) && (!TASKLET_IN_USE)) {
3718                                 DBG_PRINT(INTR_DBG, "%s: Rx BD hit ",
3719                                                         dev->name);
3720                                 DBG_PRINT(INTR_DBG, "PANIC levels\n");
3721                                 if ((ret = fill_rx_buffers(sp, i)) == -ENOMEM) {
3722                                         DBG_PRINT(ERR_DBG, "%s:Out of memory",
3723                                                   dev->name);
3724                                         DBG_PRINT(ERR_DBG, " in ISR!!\n");
3725                                         clear_bit(0, (&sp->tasklet_status));
3726                                         atomic_dec(&sp->isr_cnt);
3727                                         return IRQ_HANDLED;
3728                                 }
3729                                 clear_bit(0, (&sp->tasklet_status));
3730                         } else if (level == LOW) {
3731                                 tasklet_schedule(&sp->task);
3732                         }
3733                 }
3734                 else if (fill_rx_buffers(sp, i) == -ENOMEM) {
3735                                 DBG_PRINT(ERR_DBG, "%s:Out of memory",
3736                                                         dev->name);
3737                                 DBG_PRINT(ERR_DBG, " in Rx Intr!!\n");
3738                                 break;
3739                 }
3740         }
3741
3742         atomic_dec(&sp->isr_cnt);
3743         return IRQ_HANDLED;
3744 }
3745
3746 static irqreturn_t
3747 s2io_msix_ring_handle(int irq, void *dev_id, struct pt_regs *regs)
3748 {
3749         ring_info_t *ring = (ring_info_t *)dev_id;
3750         nic_t *sp = ring->nic;
3751         struct net_device *dev = (struct net_device *) dev_id;
3752         int rxb_size, level, rng_n;
3753
3754         atomic_inc(&sp->isr_cnt);
3755         rx_intr_handler(ring);
3756
3757         rng_n = ring->ring_no;
3758         if (!sp->lro) {
3759                 rxb_size = atomic_read(&sp->rx_bufs_left[rng_n]);
3760                 level = rx_buffer_level(sp, rxb_size, rng_n);
3761
3762                 if ((level == PANIC) && (!TASKLET_IN_USE)) {
3763                         int ret;
3764                         DBG_PRINT(INTR_DBG, "%s: Rx BD hit ", __FUNCTION__);
3765                         DBG_PRINT(INTR_DBG, "PANIC levels\n");
3766                         if ((ret = fill_rx_buffers(sp, rng_n)) == -ENOMEM) {
3767                                 DBG_PRINT(ERR_DBG, "Out of memory in %s",
3768                                           __FUNCTION__);
3769                                 clear_bit(0, (&sp->tasklet_status));
3770                                 return IRQ_HANDLED;
3771                         }
3772                         clear_bit(0, (&sp->tasklet_status));
3773                 } else if (level == LOW) {
3774                         tasklet_schedule(&sp->task);
3775                 }
3776         }
3777         else if (fill_rx_buffers(sp, rng_n) == -ENOMEM) {
3778                         DBG_PRINT(ERR_DBG, "%s:Out of memory", dev->name);
3779                         DBG_PRINT(ERR_DBG, " in Rx Intr!!\n");
3780         }
3781
3782         atomic_dec(&sp->isr_cnt);
3783
3784         return IRQ_HANDLED;
3785 }
3786
3787 static irqreturn_t
3788 s2io_msix_fifo_handle(int irq, void *dev_id, struct pt_regs *regs)
3789 {
3790         fifo_info_t *fifo = (fifo_info_t *)dev_id;
3791         nic_t *sp = fifo->nic;
3792
3793         atomic_inc(&sp->isr_cnt);
3794         tx_intr_handler(fifo);
3795         atomic_dec(&sp->isr_cnt);
3796         return IRQ_HANDLED;
3797 }
3798
3799 static void s2io_txpic_intr_handle(nic_t *sp)
3800 {
3801         XENA_dev_config_t __iomem *bar0 = sp->bar0;
3802         u64 val64;
3803
3804         val64 = readq(&bar0->pic_int_status);
3805         if (val64 & PIC_INT_GPIO) {
3806                 val64 = readq(&bar0->gpio_int_reg);
3807                 if ((val64 & GPIO_INT_REG_LINK_DOWN) &&
3808                     (val64 & GPIO_INT_REG_LINK_UP)) {
3809                         val64 |=  GPIO_INT_REG_LINK_DOWN;
3810                         val64 |= GPIO_INT_REG_LINK_UP;
3811                         writeq(val64, &bar0->gpio_int_reg);
3812                         goto masking;
3813                 }
3814
3815                 if (((sp->last_link_state == LINK_UP) &&
3816                         (val64 & GPIO_INT_REG_LINK_DOWN)) ||
3817                 ((sp->last_link_state == LINK_DOWN) &&
3818                 (val64 & GPIO_INT_REG_LINK_UP))) {
3819                         val64 = readq(&bar0->gpio_int_mask);
3820                         val64 |=  GPIO_INT_MASK_LINK_DOWN;
3821                         val64 |= GPIO_INT_MASK_LINK_UP;
3822                         writeq(val64, &bar0->gpio_int_mask);
3823                         s2io_set_link((unsigned long)sp);
3824                 }
3825 masking:
3826                 if (sp->last_link_state == LINK_UP) {
3827                         /*enable down interrupt */
3828                         val64 = readq(&bar0->gpio_int_mask);
3829                         /* unmasks link down intr */
3830                         val64 &=  ~GPIO_INT_MASK_LINK_DOWN;
3831                         /* masks link up intr */
3832                         val64 |= GPIO_INT_MASK_LINK_UP;
3833                         writeq(val64, &bar0->gpio_int_mask);
3834                 } else {
3835                         /*enable UP Interrupt */
3836                         val64 = readq(&bar0->gpio_int_mask);
3837                         /* unmasks link up interrupt */
3838                         val64 &= ~GPIO_INT_MASK_LINK_UP;
3839                         /* masks link down interrupt */
3840                         val64 |=  GPIO_INT_MASK_LINK_DOWN;
3841                         writeq(val64, &bar0->gpio_int_mask);
3842                 }
3843         }
3844 }
3845
3846 /**
3847  *  s2io_isr - ISR handler of the device .
3848  *  @irq: the irq of the device.
3849  *  @dev_id: a void pointer to the dev structure of the NIC.
3850  *  @pt_regs: pointer to the registers pushed on the stack.
3851  *  Description:  This function is the ISR handler of the device. It
3852  *  identifies the reason for the interrupt and calls the relevant
3853  *  service routines. As a contongency measure, this ISR allocates the
3854  *  recv buffers, if their numbers are below the panic value which is
3855  *  presently set to 25% of the original number of rcv buffers allocated.
3856  *  Return value:
3857  *   IRQ_HANDLED: will be returned if IRQ was handled by this routine
3858  *   IRQ_NONE: will be returned if interrupt is not from our device
3859  */
3860 static irqreturn_t s2io_isr(int irq, void *dev_id, struct pt_regs *regs)
3861 {
3862         struct net_device *dev = (struct net_device *) dev_id;
3863         nic_t *sp = dev->priv;
3864         XENA_dev_config_t __iomem *bar0 = sp->bar0;
3865         int i;
3866         u64 reason = 0, val64;
3867         mac_info_t *mac_control;
3868         struct config_param *config;
3869
3870         atomic_inc(&sp->isr_cnt);
3871         mac_control = &sp->mac_control;
3872         config = &sp->config;
3873
3874         /*
3875          * Identify the cause for interrupt and call the appropriate
3876          * interrupt handler. Causes for the interrupt could be;
3877          * 1. Rx of packet.
3878          * 2. Tx complete.
3879          * 3. Link down.
3880          * 4. Error in any functional blocks of the NIC.
3881          */
3882         reason = readq(&bar0->general_int_status);
3883
3884         if (!reason) {
3885                 /* The interrupt was not raised by Xena. */
3886                 atomic_dec(&sp->isr_cnt);
3887                 return IRQ_NONE;
3888         }
3889
3890 #ifdef CONFIG_S2IO_NAPI
3891         if (reason & GEN_INTR_RXTRAFFIC) {
3892                 if (netif_rx_schedule_prep(dev)) {
3893                         en_dis_able_nic_intrs(sp, RX_TRAFFIC_INTR,
3894                                               DISABLE_INTRS);
3895                         __netif_rx_schedule(dev);
3896                 }
3897         }
3898 #else
3899         /* If Intr is because of Rx Traffic */
3900         if (reason & GEN_INTR_RXTRAFFIC) {
3901                 /*
3902                  * rx_traffic_int reg is an R1 register, writing all 1's
3903                  * will ensure that the actual interrupt causing bit get's
3904                  * cleared and hence a read can be avoided.
3905                  */
3906                 val64 = 0xFFFFFFFFFFFFFFFFULL;
3907                 writeq(val64, &bar0->rx_traffic_int);
3908                 for (i = 0; i < config->rx_ring_num; i++) {
3909                         rx_intr_handler(&mac_control->rings[i]);
3910                 }
3911         }
3912 #endif
3913
3914         /* If Intr is because of Tx Traffic */
3915         if (reason & GEN_INTR_TXTRAFFIC) {
3916                 /*
3917                  * tx_traffic_int reg is an R1 register, writing all 1's
3918                  * will ensure that the actual interrupt causing bit get's
3919                  * cleared and hence a read can be avoided.
3920                  */
3921                 val64 = 0xFFFFFFFFFFFFFFFFULL;
3922                 writeq(val64, &bar0->tx_traffic_int);
3923
3924                 for (i = 0; i < config->tx_fifo_num; i++)
3925                         tx_intr_handler(&mac_control->fifos[i]);
3926         }
3927
3928         if (reason & GEN_INTR_TXPIC)
3929                 s2io_txpic_intr_handle(sp);
3930         /*
3931          * If the Rx buffer count is below the panic threshold then
3932          * reallocate the buffers from the interrupt handler itself,
3933          * else schedule a tasklet to reallocate the buffers.
3934          */
3935 #ifndef CONFIG_S2IO_NAPI
3936         for (i = 0; i < config->rx_ring_num; i++) {
3937                 if (!sp->lro) {
3938                         int ret;
3939                         int rxb_size = atomic_read(&sp->rx_bufs_left[i]);
3940                         int level = rx_buffer_level(sp, rxb_size, i);
3941
3942                         if ((level == PANIC) && (!TASKLET_IN_USE)) {
3943                                 DBG_PRINT(INTR_DBG, "%s: Rx BD hit ",
3944                                                         dev->name);
3945                                 DBG_PRINT(INTR_DBG, "PANIC levels\n");
3946                                 if ((ret = fill_rx_buffers(sp, i)) == -ENOMEM) {
3947                                         DBG_PRINT(ERR_DBG, "%s:Out of memory",
3948                                                   dev->name);
3949                                         DBG_PRINT(ERR_DBG, " in ISR!!\n");
3950                                         clear_bit(0, (&sp->tasklet_status));
3951                                         atomic_dec(&sp->isr_cnt);
3952                                         return IRQ_HANDLED;
3953                                 }
3954                                 clear_bit(0, (&sp->tasklet_status));
3955                         } else if (level == LOW) {
3956                                 tasklet_schedule(&sp->task);
3957                         }
3958                 }
3959                 else if (fill_rx_buffers(sp, i) == -ENOMEM) {
3960                                 DBG_PRINT(ERR_DBG, "%s:Out of memory",
3961                                                         dev->name);
3962                                 DBG_PRINT(ERR_DBG, " in Rx intr!!\n");
3963                                 break;
3964                 }
3965         }
3966 #endif
3967
3968         atomic_dec(&sp->isr_cnt);
3969         return IRQ_HANDLED;
3970 }
3971
3972 /**
3973  * s2io_updt_stats -
3974  */
3975 static void s2io_updt_stats(nic_t *sp)
3976 {
3977         XENA_dev_config_t __iomem *bar0 = sp->bar0;
3978         u64 val64;
3979         int cnt = 0;
3980
3981         if (atomic_read(&sp->card_state) == CARD_UP) {
3982                 /* Apprx 30us on a 133 MHz bus */
3983                 val64 = SET_UPDT_CLICKS(10) |
3984                         STAT_CFG_ONE_SHOT_EN | STAT_CFG_STAT_EN;
3985                 writeq(val64, &bar0->stat_cfg);
3986                 do {
3987                         udelay(100);
3988                         val64 = readq(&bar0->stat_cfg);
3989                         if (!(val64 & BIT(0)))
3990                                 break;
3991                         cnt++;
3992                         if (cnt == 5)
3993                                 break; /* Updt failed */
3994                 } while(1);
3995         }
3996 }
3997
3998 /**
3999  *  s2io_get_stats - Updates the device statistics structure.
4000  *  @dev : pointer to the device structure.
4001  *  Description:
4002  *  This function updates the device statistics structure in the s2io_nic
4003  *  structure and returns a pointer to the same.
4004  *  Return value:
4005  *  pointer to the updated net_device_stats structure.
4006  */
4007
4008 static struct net_device_stats *s2io_get_stats(struct net_device *dev)
4009 {
4010         nic_t *sp = dev->priv;
4011         mac_info_t *mac_control;
4012         struct config_param *config;
4013
4014
4015         mac_control = &sp->mac_control;
4016         config = &sp->config;
4017
4018         /* Configure Stats for immediate updt */
4019         s2io_updt_stats(sp);
4020
4021         sp->stats.tx_packets =
4022                 le32_to_cpu(mac_control->stats_info->tmac_frms);
4023         sp->stats.tx_errors =
4024                 le32_to_cpu(mac_control->stats_info->tmac_any_err_frms);
4025         sp->stats.rx_errors =
4026                 le32_to_cpu(mac_control->stats_info->rmac_drop_frms);
4027         sp->stats.multicast =
4028                 le32_to_cpu(mac_control->stats_info->rmac_vld_mcst_frms);
4029         sp->stats.rx_length_errors =
4030                 le32_to_cpu(mac_control->stats_info->rmac_long_frms);
4031
4032         return (&sp->stats);
4033 }
4034
4035 /**
4036  *  s2io_set_multicast - entry point for multicast address enable/disable.
4037  *  @dev : pointer to the device structure
4038  *  Description:
4039  *  This function is a driver entry point which gets called by the kernel
4040  *  whenever multicast addresses must be enabled/disabled. This also gets
4041  *  called to set/reset promiscuous mode. Depending on the deivce flag, we
4042  *  determine, if multicast address must be enabled or if promiscuous mode
4043  *  is to be disabled etc.
4044  *  Return value:
4045  *  void.
4046  */
4047
4048 static void s2io_set_multicast(struct net_device *dev)
4049 {
4050         int i, j, prev_cnt;
4051         struct dev_mc_list *mclist;
4052         nic_t *sp = dev->priv;
4053         XENA_dev_config_t __iomem *bar0 = sp->bar0;
4054         u64 val64 = 0, multi_mac = 0x010203040506ULL, mask =
4055             0xfeffffffffffULL;
4056         u64 dis_addr = 0xffffffffffffULL, mac_addr = 0;
4057         void __iomem *add;
4058
4059         if ((dev->flags & IFF_ALLMULTI) && (!sp->m_cast_flg)) {
4060                 /*  Enable all Multicast addresses */
4061                 writeq(RMAC_ADDR_DATA0_MEM_ADDR(multi_mac),
4062                        &bar0->rmac_addr_data0_mem);
4063                 writeq(RMAC_ADDR_DATA1_MEM_MASK(mask),
4064                        &bar0->rmac_addr_data1_mem);
4065                 val64 = RMAC_ADDR_CMD_MEM_WE |
4066                     RMAC_ADDR_CMD_MEM_STROBE_NEW_CMD |
4067                     RMAC_ADDR_CMD_MEM_OFFSET(MAC_MC_ALL_MC_ADDR_OFFSET);
4068                 writeq(val64, &bar0->rmac_addr_cmd_mem);
4069                 /* Wait till command completes */
4070                 wait_for_cmd_complete(sp);
4071
4072                 sp->m_cast_flg = 1;
4073                 sp->all_multi_pos = MAC_MC_ALL_MC_ADDR_OFFSET;
4074         } else if ((dev->flags & IFF_ALLMULTI) && (sp->m_cast_flg)) {
4075                 /*  Disable all Multicast addresses */
4076                 writeq(RMAC_ADDR_DATA0_MEM_ADDR(dis_addr),
4077                        &bar0->rmac_addr_data0_mem);
4078                 writeq(RMAC_ADDR_DATA1_MEM_MASK(0x0),
4079                        &bar0->rmac_addr_data1_mem);
4080                 val64 = RMAC_ADDR_CMD_MEM_WE |
4081                     RMAC_ADDR_CMD_MEM_STROBE_NEW_CMD |
4082                     RMAC_ADDR_CMD_MEM_OFFSET(sp->all_multi_pos);
4083                 writeq(val64, &bar0->rmac_addr_cmd_mem);
4084                 /* Wait till command completes */
4085                 wait_for_cmd_complete(sp);
4086
4087                 sp->m_cast_flg = 0;
4088                 sp->all_multi_pos = 0;
4089         }
4090
4091         if ((dev->flags & IFF_PROMISC) && (!sp->promisc_flg)) {
4092                 /*  Put the NIC into promiscuous mode */
4093                 add = &bar0->mac_cfg;
4094                 val64 = readq(&bar0->mac_cfg);
4095                 val64 |= MAC_CFG_RMAC_PROM_ENABLE;
4096
4097                 writeq(RMAC_CFG_KEY(0x4C0D), &bar0->rmac_cfg_key);
4098                 writel((u32) val64, add);
4099                 writeq(RMAC_CFG_KEY(0x4C0D), &bar0->rmac_cfg_key);
4100                 writel((u32) (val64 >> 32), (add + 4));
4101
4102                 val64 = readq(&bar0->mac_cfg);
4103                 sp->promisc_flg = 1;
4104                 DBG_PRINT(INFO_DBG, "%s: entered promiscuous mode\n",
4105                           dev->name);
4106         } else if (!(dev->flags & IFF_PROMISC) && (sp->promisc_flg)) {
4107                 /*  Remove the NIC from promiscuous mode */
4108                 add = &bar0->mac_cfg;
4109                 val64 = readq(&bar0->mac_cfg);
4110                 val64 &= ~MAC_CFG_RMAC_PROM_ENABLE;
4111
4112                 writeq(RMAC_CFG_KEY(0x4C0D), &bar0->rmac_cfg_key);
4113                 writel((u32) val64, add);
4114                 writeq(RMAC_CFG_KEY(0x4C0D), &bar0->rmac_cfg_key);
4115                 writel((u32) (val64 >> 32), (add + 4));
4116
4117                 val64 = readq(&bar0->mac_cfg);
4118                 sp->promisc_flg = 0;
4119                 DBG_PRINT(INFO_DBG, "%s: left promiscuous mode\n",
4120                           dev->name);
4121         }
4122
4123         /*  Update individual M_CAST address list */
4124         if ((!sp->m_cast_flg) && dev->mc_count) {
4125                 if (dev->mc_count >
4126                     (MAX_ADDRS_SUPPORTED - MAC_MC_ADDR_START_OFFSET - 1)) {
4127                         DBG_PRINT(ERR_DBG, "%s: No more Rx filters ",
4128                                   dev->name);
4129                         DBG_PRINT(ERR_DBG, "can be added, please enable ");
4130                         DBG_PRINT(ERR_DBG, "ALL_MULTI instead\n");
4131                         return;
4132                 }
4133
4134                 prev_cnt = sp->mc_addr_count;
4135                 sp->mc_addr_count = dev->mc_count;
4136
4137                 /* Clear out the previous list of Mc in the H/W. */
4138                 for (i = 0; i < prev_cnt; i++) {
4139                         writeq(RMAC_ADDR_DATA0_MEM_ADDR(dis_addr),
4140                                &bar0->rmac_addr_data0_mem);
4141                         writeq(RMAC_ADDR_DATA1_MEM_MASK(0ULL),
4142                                 &bar0->rmac_addr_data1_mem);
4143                         val64 = RMAC_ADDR_CMD_MEM_WE |
4144                             RMAC_ADDR_CMD_MEM_STROBE_NEW_CMD |
4145                             RMAC_ADDR_CMD_MEM_OFFSET
4146                             (MAC_MC_ADDR_START_OFFSET + i);
4147                         writeq(val64, &bar0->rmac_addr_cmd_mem);
4148
4149                         /* Wait for command completes */
4150                         if (wait_for_cmd_complete(sp)) {
4151                                 DBG_PRINT(ERR_DBG, "%s: Adding ",
4152                                           dev->name);
4153                                 DBG_PRINT(ERR_DBG, "Multicasts failed\n");
4154                                 return;
4155                         }
4156                 }
4157
4158                 /* Create the new Rx filter list and update the same in H/W. */
4159                 for (i = 0, mclist = dev->mc_list; i < dev->mc_count;
4160                      i++, mclist = mclist->next) {
4161                         memcpy(sp->usr_addrs[i].addr, mclist->dmi_addr,
4162                                ETH_ALEN);
4163                         mac_addr = 0;
4164                         for (j = 0; j < ETH_ALEN; j++) {
4165                                 mac_addr |= mclist->dmi_addr[j];
4166                                 mac_addr <<= 8;
4167                         }
4168                         mac_addr >>= 8;
4169                         writeq(RMAC_ADDR_DATA0_MEM_ADDR(mac_addr),
4170                                &bar0->rmac_addr_data0_mem);
4171                         writeq(RMAC_ADDR_DATA1_MEM_MASK(0ULL),
4172                                 &bar0->rmac_addr_data1_mem);
4173                         val64 = RMAC_ADDR_CMD_MEM_WE |
4174                             RMAC_ADDR_CMD_MEM_STROBE_NEW_CMD |
4175                             RMAC_ADDR_CMD_MEM_OFFSET
4176                             (i + MAC_MC_ADDR_START_OFFSET);
4177                         writeq(val64, &bar0->rmac_addr_cmd_mem);
4178
4179                         /* Wait for command completes */
4180                         if (wait_for_cmd_complete(sp)) {
4181                                 DBG_PRINT(ERR_DBG, "%s: Adding ",
4182                                           dev->name);
4183                                 DBG_PRINT(ERR_DBG, "Multicasts failed\n");
4184                                 return;
4185                         }
4186                 }
4187         }
4188 }
4189
4190 /**
4191  *  s2io_set_mac_addr - Programs the Xframe mac address
4192  *  @dev : pointer to the device structure.
4193  *  @addr: a uchar pointer to the new mac address which is to be set.
4194  *  Description : This procedure will program the Xframe to receive
4195  *  frames with new Mac Address
4196  *  Return value: SUCCESS on success and an appropriate (-)ve integer
4197  *  as defined in errno.h file on failure.
4198  */
4199
4200 static int s2io_set_mac_addr(struct net_device *dev, u8 * addr)
4201 {
4202         nic_t *sp = dev->priv;
4203         XENA_dev_config_t __iomem *bar0 = sp->bar0;
4204         register u64 val64, mac_addr = 0;
4205         int i;
4206
4207         /*
4208          * Set the new MAC address as the new unicast filter and reflect this
4209          * change on the device address registered with the OS. It will be
4210          * at offset 0.
4211          */
4212         for (i = 0; i < ETH_ALEN; i++) {
4213                 mac_addr <<= 8;
4214                 mac_addr |= addr[i];
4215         }
4216
4217         writeq(RMAC_ADDR_DATA0_MEM_ADDR(mac_addr),
4218                &bar0->rmac_addr_data0_mem);
4219
4220         val64 =
4221             RMAC_ADDR_CMD_MEM_WE | RMAC_ADDR_CMD_MEM_STROBE_NEW_CMD |
4222             RMAC_ADDR_CMD_MEM_OFFSET(0);
4223         writeq(val64, &bar0->rmac_addr_cmd_mem);
4224         /* Wait till command completes */
4225         if (wait_for_cmd_complete(sp)) {
4226                 DBG_PRINT(ERR_DBG, "%s: set_mac_addr failed\n", dev->name);
4227                 return FAILURE;
4228         }
4229
4230         return SUCCESS;
4231 }
4232
4233 /**
4234  * s2io_ethtool_sset - Sets different link parameters.
4235  * @sp : private member of the device structure, which is a pointer to the  * s2io_nic structure.
4236  * @info: pointer to the structure with parameters given by ethtool to set
4237  * link information.
4238  * Description:
4239  * The function sets different link parameters provided by the user onto
4240  * the NIC.
4241  * Return value:
4242  * 0 on success.
4243 */
4244
4245 static int s2io_ethtool_sset(struct net_device *dev,
4246                              struct ethtool_cmd *info)
4247 {
4248         nic_t *sp = dev->priv;
4249         if ((info->autoneg == AUTONEG_ENABLE) ||
4250             (info->speed != SPEED_10000) || (info->duplex != DUPLEX_FULL))
4251                 return -EINVAL;
4252         else {
4253                 s2io_close(sp->dev);
4254                 s2io_open(sp->dev);
4255         }
4256
4257         return 0;
4258 }
4259
4260 /**
4261  * s2io_ethtol_gset - Return link specific information.
4262  * @sp : private member of the device structure, pointer to the
4263  *      s2io_nic structure.
4264  * @info : pointer to the structure with parameters given by ethtool
4265  * to return link information.
4266  * Description:
4267  * Returns link specific information like speed, duplex etc.. to ethtool.
4268  * Return value :
4269  * return 0 on success.
4270  */
4271
4272 static int s2io_ethtool_gset(struct net_device *dev, struct ethtool_cmd *info)
4273 {
4274         nic_t *sp = dev->priv;
4275         info->supported = (SUPPORTED_10000baseT_Full | SUPPORTED_FIBRE);
4276         info->advertising = (SUPPORTED_10000baseT_Full | SUPPORTED_FIBRE);
4277         info->port = PORT_FIBRE;
4278         /* info->transceiver?? TODO */
4279
4280         if (netif_carrier_ok(sp->dev)) {
4281                 info->speed = 10000;
4282                 info->duplex = DUPLEX_FULL;
4283         } else {
4284                 info->speed = -1;
4285                 info->duplex = -1;
4286         }
4287
4288         info->autoneg = AUTONEG_DISABLE;
4289         return 0;
4290 }
4291
4292 /**
4293  * s2io_ethtool_gdrvinfo - Returns driver specific information.
4294  * @sp : private member of the device structure, which is a pointer to the
4295  * s2io_nic structure.
4296  * @info : pointer to the structure with parameters given by ethtool to
4297  * return driver information.
4298  * Description:
4299  * Returns driver specefic information like name, version etc.. to ethtool.
4300  * Return value:
4301  *  void
4302  */
4303
4304 static void s2io_ethtool_gdrvinfo(struct net_device *dev,
4305                                   struct ethtool_drvinfo *info)
4306 {
4307         nic_t *sp = dev->priv;
4308
4309         strncpy(info->driver, s2io_driver_name, sizeof(info->driver));
4310         strncpy(info->version, s2io_driver_version, sizeof(info->version));
4311         strncpy(info->fw_version, "", sizeof(info->fw_version));
4312         strncpy(info->bus_info, pci_name(sp->pdev), sizeof(info->bus_info));
4313         info->regdump_len = XENA_REG_SPACE;
4314         info->eedump_len = XENA_EEPROM_SPACE;
4315         info->testinfo_len = S2IO_TEST_LEN;
4316         info->n_stats = S2IO_STAT_LEN;
4317 }
4318
4319 /**
4320  *  s2io_ethtool_gregs - dumps the entire space of Xfame into the buffer.
4321  *  @sp: private member of the device structure, which is a pointer to the
4322  *  s2io_nic structure.
4323  *  @regs : pointer to the structure with parameters given by ethtool for
4324  *  dumping the registers.
4325  *  @reg_space: The input argumnet into which all the registers are dumped.
4326  *  Description:
4327  *  Dumps the entire register space of xFrame NIC into the user given
4328  *  buffer area.
4329  * Return value :
4330  * void .
4331 */
4332
4333 static void s2io_ethtool_gregs(struct net_device *dev,
4334                                struct ethtool_regs *regs, void *space)
4335 {
4336         int i;
4337         u64 reg;
4338         u8 *reg_space = (u8 *) space;
4339         nic_t *sp = dev->priv;
4340
4341         regs->len = XENA_REG_SPACE;
4342         regs->version = sp->pdev->subsystem_device;
4343
4344         for (i = 0; i < regs->len; i += 8) {
4345                 reg = readq(sp->bar0 + i);
4346                 memcpy((reg_space + i), &reg, 8);
4347         }
4348 }
4349
4350 /**
4351  *  s2io_phy_id  - timer function that alternates adapter LED.
4352  *  @data : address of the private member of the device structure, which
4353  *  is a pointer to the s2io_nic structure, provided as an u32.
4354  * Description: This is actually the timer function that alternates the
4355  * adapter LED bit of the adapter control bit to set/reset every time on
4356  * invocation. The timer is set for 1/2 a second, hence tha NIC blinks
4357  *  once every second.
4358 */
4359 static void s2io_phy_id(unsigned long data)
4360 {
4361         nic_t *sp = (nic_t *) data;
4362         XENA_dev_config_t __iomem *bar0 = sp->bar0;
4363         u64 val64 = 0;
4364         u16 subid;
4365
4366         subid = sp->pdev->subsystem_device;
4367         if ((sp->device_type == XFRAME_II_DEVICE) ||
4368                    ((subid & 0xFF) >= 0x07)) {
4369                 val64 = readq(&bar0->gpio_control);
4370                 val64 ^= GPIO_CTRL_GPIO_0;
4371                 writeq(val64, &bar0->gpio_control);
4372         } else {
4373                 val64 = readq(&bar0->adapter_control);
4374                 val64 ^= ADAPTER_LED_ON;
4375                 writeq(val64, &bar0->adapter_control);
4376         }
4377
4378         mod_timer(&sp->id_timer, jiffies + HZ / 2);
4379 }
4380
4381 /**
4382  * s2io_ethtool_idnic - To physically identify the nic on the system.
4383  * @sp : private member of the device structure, which is a pointer to the
4384  * s2io_nic structure.
4385  * @id : pointer to the structure with identification parameters given by
4386  * ethtool.
4387  * Description: Used to physically identify the NIC on the system.
4388  * The Link LED will blink for a time specified by the user for
4389  * identification.
4390  * NOTE: The Link has to be Up to be able to blink the LED. Hence
4391  * identification is possible only if it's link is up.
4392  * Return value:
4393  * int , returns 0 on success
4394  */
4395
4396 static int s2io_ethtool_idnic(struct net_device *dev, u32 data)
4397 {
4398         u64 val64 = 0, last_gpio_ctrl_val;
4399         nic_t *sp = dev->priv;
4400         XENA_dev_config_t __iomem *bar0 = sp->bar0;
4401         u16 subid;
4402
4403         subid = sp->pdev->subsystem_device;
4404         last_gpio_ctrl_val = readq(&bar0->gpio_control);
4405         if ((sp->device_type == XFRAME_I_DEVICE) &&
4406                 ((subid & 0xFF) < 0x07)) {
4407                 val64 = readq(&bar0->adapter_control);
4408                 if (!(val64 & ADAPTER_CNTL_EN)) {
4409                         printk(KERN_ERR
4410                                "Adapter Link down, cannot blink LED\n");
4411                         return -EFAULT;
4412                 }
4413         }
4414         if (sp->id_timer.function == NULL) {
4415                 init_timer(&sp->id_timer);
4416                 sp->id_timer.function = s2io_phy_id;
4417                 sp->id_timer.data = (unsigned long) sp;
4418         }
4419         mod_timer(&sp->id_timer, jiffies);
4420         if (data)
4421                 msleep_interruptible(data * HZ);
4422         else
4423                 msleep_interruptible(MAX_FLICKER_TIME);
4424         del_timer_sync(&sp->id_timer);
4425
4426         if (CARDS_WITH_FAULTY_LINK_INDICATORS(sp->device_type, subid)) {
4427                 writeq(last_gpio_ctrl_val, &bar0->gpio_control);
4428                 last_gpio_ctrl_val = readq(&bar0->gpio_control);
4429         }
4430
4431         return 0;
4432 }
4433
4434 /**
4435  * s2io_ethtool_getpause_data -Pause frame frame generation and reception.
4436  * @sp : private member of the device structure, which is a pointer to the
4437  *      s2io_nic structure.
4438  * @ep : pointer to the structure with pause parameters given by ethtool.
4439  * Description:
4440  * Returns the Pause frame generation and reception capability of the NIC.
4441  * Return value:
4442  *  void
4443  */
4444 static void s2io_ethtool_getpause_data(struct net_device *dev,
4445                                        struct ethtool_pauseparam *ep)
4446 {
4447         u64 val64;
4448         nic_t *sp = dev->priv;
4449         XENA_dev_config_t __iomem *bar0 = sp->bar0;
4450
4451         val64 = readq(&bar0->rmac_pause_cfg);
4452         if (val64 & RMAC_PAUSE_GEN_ENABLE)
4453                 ep->tx_pause = TRUE;
4454         if (val64 & RMAC_PAUSE_RX_ENABLE)
4455                 ep->rx_pause = TRUE;
4456         ep->autoneg = FALSE;
4457 }
4458
4459 /**
4460  * s2io_ethtool_setpause_data -  set/reset pause frame generation.
4461  * @sp : private member of the device structure, which is a pointer to the
4462  *      s2io_nic structure.
4463  * @ep : pointer to the structure with pause parameters given by ethtool.
4464  * Description:
4465  * It can be used to set or reset Pause frame generation or reception
4466  * support of the NIC.
4467  * Return value:
4468  * int, returns 0 on Success
4469  */
4470
4471 static int s2io_ethtool_setpause_data(struct net_device *dev,
4472                                struct ethtool_pauseparam *ep)
4473 {
4474         u64 val64;
4475         nic_t *sp = dev->priv;
4476         XENA_dev_config_t __iomem *bar0 = sp->bar0;
4477
4478         val64 = readq(&bar0->rmac_pause_cfg);
4479         if (ep->tx_pause)
4480                 val64 |= RMAC_PAUSE_GEN_ENABLE;
4481         else
4482                 val64 &= ~RMAC_PAUSE_GEN_ENABLE;
4483         if (ep->rx_pause)
4484                 val64 |= RMAC_PAUSE_RX_ENABLE;
4485         else
4486                 val64 &= ~RMAC_PAUSE_RX_ENABLE;
4487         writeq(val64, &bar0->rmac_pause_cfg);
4488         return 0;
4489 }
4490
4491 /**
4492  * read_eeprom - reads 4 bytes of data from user given offset.
4493  * @sp : private member of the device structure, which is a pointer to the
4494  *      s2io_nic structure.
4495  * @off : offset at which the data must be written
4496  * @data : Its an output parameter where the data read at the given
4497  *      offset is stored.
4498  * Description:
4499  * Will read 4 bytes of data from the user given offset and return the
4500  * read data.
4501  * NOTE: Will allow to read only part of the EEPROM visible through the
4502  *   I2C bus.
4503  * Return value:
4504  *  -1 on failure and 0 on success.
4505  */
4506
4507 #define S2IO_DEV_ID             5
4508 static int read_eeprom(nic_t * sp, int off, u64 * data)
4509 {
4510         int ret = -1;
4511         u32 exit_cnt = 0;
4512         u64 val64;
4513         XENA_dev_config_t __iomem *bar0 = sp->bar0;
4514
4515         if (sp->device_type == XFRAME_I_DEVICE) {
4516                 val64 = I2C_CONTROL_DEV_ID(S2IO_DEV_ID) | I2C_CONTROL_ADDR(off) |
4517                     I2C_CONTROL_BYTE_CNT(0x3) | I2C_CONTROL_READ |
4518                     I2C_CONTROL_CNTL_START;
4519                 SPECIAL_REG_WRITE(val64, &bar0->i2c_control, LF);
4520
4521                 while (exit_cnt < 5) {
4522                         val64 = readq(&bar0->i2c_control);
4523                         if (I2C_CONTROL_CNTL_END(val64)) {
4524                                 *data = I2C_CONTROL_GET_DATA(val64);
4525                                 ret = 0;
4526                                 break;
4527                         }
4528                         msleep(50);
4529                         exit_cnt++;
4530                 }
4531         }
4532
4533         if (sp->device_type == XFRAME_II_DEVICE) {
4534                 val64 = SPI_CONTROL_KEY(0x9) | SPI_CONTROL_SEL1 |
4535                         SPI_CONTROL_BYTECNT(0x3) |
4536                         SPI_CONTROL_CMD(0x3) | SPI_CONTROL_ADDR(off);
4537                 SPECIAL_REG_WRITE(val64, &bar0->spi_control, LF);
4538                 val64 |= SPI_CONTROL_REQ;
4539                 SPECIAL_REG_WRITE(val64, &bar0->spi_control, LF);
4540                 while (exit_cnt < 5) {
4541                         val64 = readq(&bar0->spi_control);
4542                         if (val64 & SPI_CONTROL_NACK) {
4543                                 ret = 1;
4544                                 break;
4545                         } else if (val64 & SPI_CONTROL_DONE) {
4546                                 *data = readq(&bar0->spi_data);
4547                                 *data &= 0xffffff;
4548                                 ret = 0;
4549                                 break;
4550                         }
4551                         msleep(50);
4552                         exit_cnt++;
4553                 }
4554         }
4555         return ret;
4556 }
4557
4558 /**
4559  *  write_eeprom - actually writes the relevant part of the data value.
4560  *  @sp : private member of the device structure, which is a pointer to the
4561  *       s2io_nic structure.
4562  *  @off : offset at which the data must be written
4563  *  @data : The data that is to be written
4564  *  @cnt : Number of bytes of the data that are actually to be written into
4565  *  the Eeprom. (max of 3)
4566  * Description:
4567  *  Actually writes the relevant part of the data value into the Eeprom
4568  *  through the I2C bus.
4569  * Return value:
4570  *  0 on success, -1 on failure.
4571  */
4572
4573 static int write_eeprom(nic_t * sp, int off, u64 data, int cnt)
4574 {
4575         int exit_cnt = 0, ret = -1;
4576         u64 val64;
4577         XENA_dev_config_t __iomem *bar0 = sp->bar0;
4578
4579         if (sp->device_type == XFRAME_I_DEVICE) {
4580                 val64 = I2C_CONTROL_DEV_ID(S2IO_DEV_ID) | I2C_CONTROL_ADDR(off) |
4581                     I2C_CONTROL_BYTE_CNT(cnt) | I2C_CONTROL_SET_DATA((u32)data) |
4582                     I2C_CONTROL_CNTL_START;
4583                 SPECIAL_REG_WRITE(val64, &bar0->i2c_control, LF);
4584
4585                 while (exit_cnt < 5) {
4586                         val64 = readq(&bar0->i2c_control);
4587                         if (I2C_CONTROL_CNTL_END(val64)) {
4588                                 if (!(val64 & I2C_CONTROL_NACK))
4589                                         ret = 0;
4590                                 break;
4591                         }
4592                         msleep(50);
4593                         exit_cnt++;
4594                 }
4595         }
4596
4597         if (sp->device_type == XFRAME_II_DEVICE) {
4598                 int write_cnt = (cnt == 8) ? 0 : cnt;
4599                 writeq(SPI_DATA_WRITE(data,(cnt<<3)), &bar0->spi_data);
4600
4601                 val64 = SPI_CONTROL_KEY(0x9) | SPI_CONTROL_SEL1 |
4602                         SPI_CONTROL_BYTECNT(write_cnt) |
4603                         SPI_CONTROL_CMD(0x2) | SPI_CONTROL_ADDR(off);
4604                 SPECIAL_REG_WRITE(val64, &bar0->spi_control, LF);
4605                 val64 |= SPI_CONTROL_REQ;
4606                 SPECIAL_REG_WRITE(val64, &bar0->spi_control, LF);
4607                 while (exit_cnt < 5) {
4608                         val64 = readq(&bar0->spi_control);
4609                         if (val64 & SPI_CONTROL_NACK) {
4610                                 ret = 1;
4611                                 break;
4612                         } else if (val64 & SPI_CONTROL_DONE) {
4613                                 ret = 0;
4614                                 break;
4615                         }
4616                         msleep(50);
4617                         exit_cnt++;
4618                 }
4619         }
4620         return ret;
4621 }
4622
4623 /**
4624  *  s2io_ethtool_geeprom  - reads the value stored in the Eeprom.
4625  *  @sp : private member of the device structure, which is a pointer to the *       s2io_nic structure.
4626  *  @eeprom : pointer to the user level structure provided by ethtool,
4627  *  containing all relevant information.
4628  *  @data_buf : user defined value to be written into Eeprom.
4629  *  Description: Reads the values stored in the Eeprom at given offset
4630  *  for a given length. Stores these values int the input argument data
4631  *  buffer 'data_buf' and returns these to the caller (ethtool.)
4632  *  Return value:
4633  *  int  0 on success
4634  */
4635
4636 static int s2io_ethtool_geeprom(struct net_device *dev,
4637                          struct ethtool_eeprom *eeprom, u8 * data_buf)
4638 {
4639         u32 i, valid;
4640         u64 data;
4641         nic_t *sp = dev->priv;
4642
4643         eeprom->magic = sp->pdev->vendor | (sp->pdev->device << 16);
4644
4645         if ((eeprom->offset + eeprom->len) > (XENA_EEPROM_SPACE))
4646                 eeprom->len = XENA_EEPROM_SPACE - eeprom->offset;
4647
4648         for (i = 0; i < eeprom->len; i += 4) {
4649                 if (read_eeprom(sp, (eeprom->offset + i), &data)) {
4650                         DBG_PRINT(ERR_DBG, "Read of EEPROM failed\n");
4651                         return -EFAULT;
4652                 }
4653                 valid = INV(data);
4654                 memcpy((data_buf + i), &valid, 4);
4655         }
4656         return 0;
4657 }
4658
4659 /**
4660  *  s2io_ethtool_seeprom - tries to write the user provided value in Eeprom
4661  *  @sp : private member of the device structure, which is a pointer to the
4662  *  s2io_nic structure.
4663  *  @eeprom : pointer to the user level structure provided by ethtool,
4664  *  containing all relevant information.
4665  *  @data_buf ; user defined value to be written into Eeprom.
4666  *  Description:
4667  *  Tries to write the user provided value in the Eeprom, at the offset
4668  *  given by the user.
4669  *  Return value:
4670  *  0 on success, -EFAULT on failure.
4671  */
4672
4673 static int s2io_ethtool_seeprom(struct net_device *dev,
4674                                 struct ethtool_eeprom *eeprom,
4675                                 u8 * data_buf)
4676 {
4677         int len = eeprom->len, cnt = 0;
4678         u64 valid = 0, data;
4679         nic_t *sp = dev->priv;
4680
4681         if (eeprom->magic != (sp->pdev->vendor | (sp->pdev->device << 16))) {
4682                 DBG_PRINT(ERR_DBG,
4683                           "ETHTOOL_WRITE_EEPROM Err: Magic value ");
4684                 DBG_PRINT(ERR_DBG, "is wrong, Its not 0x%x\n",
4685                           eeprom->magic);
4686                 return -EFAULT;
4687         }
4688
4689         while (len) {
4690                 data = (u32) data_buf[cnt] & 0x000000FF;
4691                 if (data) {
4692                         valid = (u32) (data << 24);
4693                 } else
4694                         valid = data;
4695
4696                 if (write_eeprom(sp, (eeprom->offset + cnt), valid, 0)) {
4697                         DBG_PRINT(ERR_DBG,
4698                                   "ETHTOOL_WRITE_EEPROM Err: Cannot ");
4699                         DBG_PRINT(ERR_DBG,
4700                                   "write into the specified offset\n");
4701                         return -EFAULT;
4702                 }
4703                 cnt++;
4704                 len--;
4705         }
4706
4707         return 0;
4708 }
4709
4710 /**
4711  * s2io_register_test - reads and writes into all clock domains.
4712  * @sp : private member of the device structure, which is a pointer to the
4713  * s2io_nic structure.
4714  * @data : variable that returns the result of each of the test conducted b
4715  * by the driver.
4716  * Description:
4717  * Read and write into all clock domains. The NIC has 3 clock domains,
4718  * see that registers in all the three regions are accessible.
4719  * Return value:
4720  * 0 on success.
4721  */
4722
4723 static int s2io_register_test(nic_t * sp, uint64_t * data)
4724 {
4725         XENA_dev_config_t __iomem *bar0 = sp->bar0;
4726         u64 val64 = 0, exp_val;
4727         int fail = 0;
4728
4729         val64 = readq(&bar0->pif_rd_swapper_fb);
4730         if (val64 != 0x123456789abcdefULL) {
4731                 fail = 1;
4732                 DBG_PRINT(INFO_DBG, "Read Test level 1 fails\n");
4733         }
4734
4735         val64 = readq(&bar0->rmac_pause_cfg);
4736         if (val64 != 0xc000ffff00000000ULL) {
4737                 fail = 1;
4738                 DBG_PRINT(INFO_DBG, "Read Test level 2 fails\n");
4739         }
4740
4741         val64 = readq(&bar0->rx_queue_cfg);
4742         if (sp->device_type == XFRAME_II_DEVICE)
4743                 exp_val = 0x0404040404040404ULL;
4744         else
4745                 exp_val = 0x0808080808080808ULL;
4746         if (val64 != exp_val) {
4747                 fail = 1;
4748                 DBG_PRINT(INFO_DBG, "Read Test level 3 fails\n");
4749         }
4750
4751         val64 = readq(&bar0->xgxs_efifo_cfg);
4752         if (val64 != 0x000000001923141EULL) {
4753                 fail = 1;
4754                 DBG_PRINT(INFO_DBG, "Read Test level 4 fails\n");
4755         }
4756
4757         val64 = 0x5A5A5A5A5A5A5A5AULL;
4758         writeq(val64, &bar0->xmsi_data);
4759         val64 = readq(&bar0->xmsi_data);
4760         if (val64 != 0x5A5A5A5A5A5A5A5AULL) {
4761                 fail = 1;
4762                 DBG_PRINT(ERR_DBG, "Write Test level 1 fails\n");
4763         }
4764
4765         val64 = 0xA5A5A5A5A5A5A5A5ULL;
4766         writeq(val64, &bar0->xmsi_data);
4767         val64 = readq(&bar0->xmsi_data);
4768         if (val64 != 0xA5A5A5A5A5A5A5A5ULL) {
4769                 fail = 1;
4770                 DBG_PRINT(ERR_DBG, "Write Test level 2 fails\n");
4771         }
4772
4773         *data = fail;
4774         return fail;
4775 }
4776
4777 /**
4778  * s2io_eeprom_test - to verify that EEprom in the xena can be programmed.
4779  * @sp : private member of the device structure, which is a pointer to the
4780  * s2io_nic structure.
4781  * @data:variable that returns the result of each of the test conducted by
4782  * the driver.
4783  * Description:
4784  * Verify that EEPROM in the xena can be programmed using I2C_CONTROL
4785  * register.
4786  * Return value:
4787  * 0 on success.
4788  */
4789
4790 static int s2io_eeprom_test(nic_t * sp, uint64_t * data)
4791 {
4792         int fail = 0;
4793         u64 ret_data, org_4F0, org_7F0;
4794         u8 saved_4F0 = 0, saved_7F0 = 0;
4795         struct net_device *dev = sp->dev;
4796
4797         /* Test Write Error at offset 0 */
4798         /* Note that SPI interface allows write access to all areas
4799          * of EEPROM. Hence doing all negative testing only for Xframe I.
4800          */
4801         if (sp->device_type == XFRAME_I_DEVICE)
4802                 if (!write_eeprom(sp, 0, 0, 3))
4803                         fail = 1;
4804
4805         /* Save current values at offsets 0x4F0 and 0x7F0 */
4806         if (!read_eeprom(sp, 0x4F0, &org_4F0))
4807                 saved_4F0 = 1;
4808         if (!read_eeprom(sp, 0x7F0, &org_7F0))
4809                 saved_7F0 = 1;
4810
4811         /* Test Write at offset 4f0 */
4812         if (write_eeprom(sp, 0x4F0, 0x012345, 3))
4813                 fail = 1;
4814         if (read_eeprom(sp, 0x4F0, &ret_data))
4815                 fail = 1;
4816
4817         if (ret_data != 0x012345) {
4818                 DBG_PRINT(ERR_DBG, "%s: eeprom test error at offset 0x4F0. "
4819                         "Data written %llx Data read %llx\n",
4820                         dev->name, (unsigned long long)0x12345,
4821                         (unsigned long long)ret_data);
4822                 fail = 1;
4823         }
4824
4825         /* Reset the EEPROM data go FFFF */
4826         write_eeprom(sp, 0x4F0, 0xFFFFFF, 3);
4827
4828         /* Test Write Request Error at offset 0x7c */
4829         if (sp->device_type == XFRAME_I_DEVICE)
4830                 if (!write_eeprom(sp, 0x07C, 0, 3))
4831                         fail = 1;
4832
4833         /* Test Write Request at offset 0x7f0 */
4834         if (write_eeprom(sp, 0x7F0, 0x012345, 3))
4835                 fail = 1;
4836         if (read_eeprom(sp, 0x7F0, &ret_data))
4837                 fail = 1;
4838
4839         if (ret_data != 0x012345) {
4840                 DBG_PRINT(ERR_DBG, "%s: eeprom test error at offset 0x7F0. "
4841                         "Data written %llx Data read %llx\n",
4842                         dev->name, (unsigned long long)0x12345,
4843                         (unsigned long long)ret_data);
4844                 fail = 1;
4845         }
4846
4847         /* Reset the EEPROM data go FFFF */
4848         write_eeprom(sp, 0x7F0, 0xFFFFFF, 3);
4849
4850         if (sp->device_type == XFRAME_I_DEVICE) {
4851                 /* Test Write Error at offset 0x80 */
4852                 if (!write_eeprom(sp, 0x080, 0, 3))
4853                         fail = 1;
4854
4855                 /* Test Write Error at offset 0xfc */
4856                 if (!write_eeprom(sp, 0x0FC, 0, 3))
4857                         fail = 1;
4858
4859                 /* Test Write Error at offset 0x100 */
4860                 if (!write_eeprom(sp, 0x100, 0, 3))
4861                         fail = 1;
4862
4863                 /* Test Write Error at offset 4ec */
4864                 if (!write_eeprom(sp, 0x4EC, 0, 3))
4865                         fail = 1;
4866         }
4867
4868         /* Restore values at offsets 0x4F0 and 0x7F0 */
4869         if (saved_4F0)
4870                 write_eeprom(sp, 0x4F0, org_4F0, 3);
4871         if (saved_7F0)
4872                 write_eeprom(sp, 0x7F0, org_7F0, 3);
4873
4874         *data = fail;
4875         return fail;
4876 }
4877
4878 /**
4879  * s2io_bist_test - invokes the MemBist test of the card .
4880  * @sp : private member of the device structure, which is a pointer to the
4881  * s2io_nic structure.
4882  * @data:variable that returns the result of each of the test conducted by
4883  * the driver.
4884  * Description:
4885  * This invokes the MemBist test of the card. We give around
4886  * 2 secs time for the Test to complete. If it's still not complete
4887  * within this peiod, we consider that the test failed.
4888  * Return value:
4889  * 0 on success and -1 on failure.
4890  */
4891
4892 static int s2io_bist_test(nic_t * sp, uint64_t * data)
4893 {
4894         u8 bist = 0;
4895         int cnt = 0, ret = -1;
4896
4897         pci_read_config_byte(sp->pdev, PCI_BIST, &bist);
4898         bist |= PCI_BIST_START;
4899         pci_write_config_word(sp->pdev, PCI_BIST, bist);
4900
4901         while (cnt < 20) {
4902                 pci_read_config_byte(sp->pdev, PCI_BIST, &bist);
4903                 if (!(bist & PCI_BIST_START)) {
4904                         *data = (bist & PCI_BIST_CODE_MASK);
4905                         ret = 0;
4906                         break;
4907                 }
4908                 msleep(100);
4909                 cnt++;
4910         }
4911
4912         return ret;
4913 }
4914
4915 /**
4916  * s2io-link_test - verifies the link state of the nic
4917  * @sp ; private member of the device structure, which is a pointer to the
4918  * s2io_nic structure.
4919  * @data: variable that returns the result of each of the test conducted by
4920  * the driver.
4921  * Description:
4922  * The function verifies the link state of the NIC and updates the input
4923  * argument 'data' appropriately.
4924  * Return value:
4925  * 0 on success.
4926  */
4927
4928 static int s2io_link_test(nic_t * sp, uint64_t * data)
4929 {
4930         XENA_dev_config_t __iomem *bar0 = sp->bar0;
4931         u64 val64;
4932
4933         val64 = readq(&bar0->adapter_status);
4934         if (val64 & ADAPTER_STATUS_RMAC_LOCAL_FAULT)
4935                 *data = 1;
4936
4937         return 0;
4938 }
4939
4940 /**
4941  * s2io_rldram_test - offline test for access to the RldRam chip on the NIC
4942  * @sp - private member of the device structure, which is a pointer to the
4943  * s2io_nic structure.
4944  * @data - variable that returns the result of each of the test
4945  * conducted by the driver.
4946  * Description:
4947  *  This is one of the offline test that tests the read and write
4948  *  access to the RldRam chip on the NIC.
4949  * Return value:
4950  *  0 on success.
4951  */
4952
4953 static int s2io_rldram_test(nic_t * sp, uint64_t * data)
4954 {
4955         XENA_dev_config_t __iomem *bar0 = sp->bar0;
4956         u64 val64;
4957         int cnt, iteration = 0, test_fail = 0;
4958
4959         val64 = readq(&bar0->adapter_control);
4960         val64 &= ~ADAPTER_ECC_EN;
4961         writeq(val64, &bar0->adapter_control);
4962
4963         val64 = readq(&bar0->mc_rldram_test_ctrl);
4964         val64 |= MC_RLDRAM_TEST_MODE;
4965         SPECIAL_REG_WRITE(val64, &bar0->mc_rldram_test_ctrl, LF);
4966
4967         val64 = readq(&bar0->mc_rldram_mrs);
4968         val64 |= MC_RLDRAM_QUEUE_SIZE_ENABLE;
4969         SPECIAL_REG_WRITE(val64, &bar0->mc_rldram_mrs, UF);
4970
4971         val64 |= MC_RLDRAM_MRS_ENABLE;
4972         SPECIAL_REG_WRITE(val64, &bar0->mc_rldram_mrs, UF);
4973
4974         while (iteration < 2) {
4975                 val64 = 0x55555555aaaa0000ULL;
4976                 if (iteration == 1) {
4977                         val64 ^= 0xFFFFFFFFFFFF0000ULL;
4978                 }
4979                 writeq(val64, &bar0->mc_rldram_test_d0);
4980
4981                 val64 = 0xaaaa5a5555550000ULL;
4982                 if (iteration == 1) {
4983                         val64 ^= 0xFFFFFFFFFFFF0000ULL;
4984                 }
4985                 writeq(val64, &bar0->mc_rldram_test_d1);
4986
4987                 val64 = 0x55aaaaaaaa5a0000ULL;
4988                 if (iteration == 1) {
4989                         val64 ^= 0xFFFFFFFFFFFF0000ULL;
4990                 }
4991                 writeq(val64, &bar0->mc_rldram_test_d2);
4992
4993                 val64 = (u64) (0x0000003ffffe0100ULL);
4994                 writeq(val64, &bar0->mc_rldram_test_add);
4995
4996                 val64 = MC_RLDRAM_TEST_MODE | MC_RLDRAM_TEST_WRITE |
4997                         MC_RLDRAM_TEST_GO;
4998                 SPECIAL_REG_WRITE(val64, &bar0->mc_rldram_test_ctrl, LF);
4999
5000                 for (cnt = 0; cnt < 5; cnt++) {
5001                         val64 = readq(&bar0->mc_rldram_test_ctrl);
5002                         if (val64 & MC_RLDRAM_TEST_DONE)
5003                                 break;
5004                         msleep(200);
5005                 }
5006
5007                 if (cnt == 5)
5008                         break;
5009
5010                 val64 = MC_RLDRAM_TEST_MODE | MC_RLDRAM_TEST_GO;
5011                 SPECIAL_REG_WRITE(val64, &bar0->mc_rldram_test_ctrl, LF);
5012
5013                 for (cnt = 0; cnt < 5; cnt++) {
5014                         val64 = readq(&bar0->mc_rldram_test_ctrl);
5015                         if (val64 & MC_RLDRAM_TEST_DONE)
5016                                 break;
5017                         msleep(500);
5018                 }
5019
5020                 if (cnt == 5)
5021                         break;
5022
5023                 val64 = readq(&bar0->mc_rldram_test_ctrl);
5024                 if (!(val64 & MC_RLDRAM_TEST_PASS))
5025                         test_fail = 1;
5026
5027                 iteration++;
5028         }
5029
5030         *data = test_fail;
5031
5032         /* Bring the adapter out of test mode */
5033         SPECIAL_REG_WRITE(0, &bar0->mc_rldram_test_ctrl, LF);
5034
5035         return test_fail;
5036 }
5037
5038 /**
5039  *  s2io_ethtool_test - conducts 6 tsets to determine the health of card.
5040  *  @sp : private member of the device structure, which is a pointer to the
5041  *  s2io_nic structure.
5042  *  @ethtest : pointer to a ethtool command specific structure that will be
5043  *  returned to the user.
5044  *  @data : variable that returns the result of each of the test
5045  * conducted by the driver.
5046  * Description:
5047  *  This function conducts 6 tests ( 4 offline and 2 online) to determine
5048  *  the health of the card.
5049  * Return value:
5050  *  void
5051  */
5052
5053 static void s2io_ethtool_test(struct net_device *dev,
5054                               struct ethtool_test *ethtest,
5055                               uint64_t * data)
5056 {
5057         nic_t *sp = dev->priv;
5058         int orig_state = netif_running(sp->dev);
5059
5060         if (ethtest->flags == ETH_TEST_FL_OFFLINE) {
5061                 /* Offline Tests. */
5062                 if (orig_state)
5063                         s2io_close(sp->dev);
5064
5065                 if (s2io_register_test(sp, &data[0]))
5066                         ethtest->flags |= ETH_TEST_FL_FAILED;
5067
5068                 s2io_reset(sp);
5069
5070                 if (s2io_rldram_test(sp, &data[3]))
5071                         ethtest->flags |= ETH_TEST_FL_FAILED;
5072
5073                 s2io_reset(sp);
5074
5075                 if (s2io_eeprom_test(sp, &data[1]))
5076                         ethtest->flags |= ETH_TEST_FL_FAILED;
5077
5078                 if (s2io_bist_test(sp, &data[4]))
5079                         ethtest->flags |= ETH_TEST_FL_FAILED;
5080
5081                 if (orig_state)
5082                         s2io_open(sp->dev);
5083
5084                 data[2] = 0;
5085         } else {
5086                 /* Online Tests. */
5087                 if (!orig_state) {
5088                         DBG_PRINT(ERR_DBG,
5089                                   "%s: is not up, cannot run test\n",
5090                                   dev->name);
5091                         data[0] = -1;
5092                         data[1] = -1;
5093                         data[2] = -1;
5094                         data[3] = -1;
5095                         data[4] = -1;
5096                 }
5097
5098                 if (s2io_link_test(sp, &data[2]))
5099                         ethtest->flags |= ETH_TEST_FL_FAILED;
5100
5101                 data[0] = 0;
5102                 data[1] = 0;
5103                 data[3] = 0;
5104                 data[4] = 0;
5105         }
5106 }
5107
5108 static void s2io_get_ethtool_stats(struct net_device *dev,
5109                                    struct ethtool_stats *estats,
5110                                    u64 * tmp_stats)
5111 {
5112         int i = 0;
5113         nic_t *sp = dev->priv;
5114         StatInfo_t *stat_info = sp->mac_control.stats_info;
5115         u64 tmp;
5116
5117         s2io_updt_stats(sp);
5118         tmp_stats[i++] =
5119                 (u64)le32_to_cpu(stat_info->tmac_frms_oflow) << 32  |
5120                 le32_to_cpu(stat_info->tmac_frms);
5121         tmp_stats[i++] =
5122                 (u64)le32_to_cpu(stat_info->tmac_data_octets_oflow) << 32 |
5123                 le32_to_cpu(stat_info->tmac_data_octets);
5124         tmp_stats[i++] = le64_to_cpu(stat_info->tmac_drop_frms);
5125         tmp_stats[i++] =
5126                 (u64)le32_to_cpu(stat_info->tmac_mcst_frms_oflow) << 32 |
5127                 le32_to_cpu(stat_info->tmac_mcst_frms);
5128         tmp_stats[i++] =
5129                 (u64)le32_to_cpu(stat_info->tmac_bcst_frms_oflow) << 32 |
5130                 le32_to_cpu(stat_info->tmac_bcst_frms);
5131         tmp_stats[i++] = le64_to_cpu(stat_info->tmac_pause_ctrl_frms);
5132         tmp_stats[i++] =
5133                 (u64)le32_to_cpu(stat_info->tmac_any_err_frms_oflow) << 32 |
5134                 le32_to_cpu(stat_info->tmac_any_err_frms);
5135         tmp_stats[i++] = le64_to_cpu(stat_info->tmac_vld_ip_octets);
5136         tmp_stats[i++] =
5137                 (u64)le32_to_cpu(stat_info->tmac_vld_ip_oflow) << 32 |
5138                 le32_to_cpu(stat_info->tmac_vld_ip);
5139         tmp_stats[i++] =
5140                 (u64)le32_to_cpu(stat_info->tmac_drop_ip_oflow) << 32 |
5141                 le32_to_cpu(stat_info->tmac_drop_ip);
5142         tmp_stats[i++] =
5143                 (u64)le32_to_cpu(stat_info->tmac_icmp_oflow) << 32 |
5144                 le32_to_cpu(stat_info->tmac_icmp);
5145         tmp_stats[i++] =
5146                 (u64)le32_to_cpu(stat_info->tmac_rst_tcp_oflow) << 32 |
5147                 le32_to_cpu(stat_info->tmac_rst_tcp);
5148         tmp_stats[i++] = le64_to_cpu(stat_info->tmac_tcp);
5149         tmp_stats[i++] = (u64)le32_to_cpu(stat_info->tmac_udp_oflow) << 32 |
5150                 le32_to_cpu(stat_info->tmac_udp);
5151         tmp_stats[i++] =
5152                 (u64)le32_to_cpu(stat_info->rmac_vld_frms_oflow) << 32 |
5153                 le32_to_cpu(stat_info->rmac_vld_frms);
5154         tmp_stats[i++] =
5155                 (u64)le32_to_cpu(stat_info->rmac_data_octets_oflow) << 32 |
5156                 le32_to_cpu(stat_info->rmac_data_octets);
5157         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_fcs_err_frms);
5158         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_drop_frms);
5159         tmp_stats[i++] =
5160                 (u64)le32_to_cpu(stat_info->rmac_vld_mcst_frms_oflow) << 32 |
5161                 le32_to_cpu(stat_info->rmac_vld_mcst_frms);
5162         tmp_stats[i++] =
5163                 (u64)le32_to_cpu(stat_info->rmac_vld_bcst_frms_oflow) << 32 |
5164                 le32_to_cpu(stat_info->rmac_vld_bcst_frms);
5165         tmp_stats[i++] = le32_to_cpu(stat_info->rmac_in_rng_len_err_frms);
5166         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_long_frms);
5167         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_pause_ctrl_frms);
5168         tmp_stats[i++] =
5169                 (u64)le32_to_cpu(stat_info->rmac_discarded_frms_oflow) << 32 |
5170                 le32_to_cpu(stat_info->rmac_discarded_frms);
5171         tmp_stats[i++] =
5172                 (u64)le32_to_cpu(stat_info->rmac_usized_frms_oflow) << 32 |
5173                 le32_to_cpu(stat_info->rmac_usized_frms);
5174         tmp_stats[i++] =
5175                 (u64)le32_to_cpu(stat_info->rmac_osized_frms_oflow) << 32 |
5176                 le32_to_cpu(stat_info->rmac_osized_frms);
5177         tmp_stats[i++] =
5178                 (u64)le32_to_cpu(stat_info->rmac_frag_frms_oflow) << 32 |
5179                 le32_to_cpu(stat_info->rmac_frag_frms);
5180         tmp_stats[i++] =
5181                 (u64)le32_to_cpu(stat_info->rmac_jabber_frms_oflow) << 32 |
5182                 le32_to_cpu(stat_info->rmac_jabber_frms);
5183         tmp_stats[i++] = (u64)le32_to_cpu(stat_info->rmac_ip_oflow) << 32 |
5184                 le32_to_cpu(stat_info->rmac_ip);
5185         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_ip_octets);
5186         tmp_stats[i++] = le32_to_cpu(stat_info->rmac_hdr_err_ip);
5187         tmp_stats[i++] = (u64)le32_to_cpu(stat_info->rmac_drop_ip_oflow) << 32 |
5188                 le32_to_cpu(stat_info->rmac_drop_ip);
5189         tmp_stats[i++] = (u64)le32_to_cpu(stat_info->rmac_icmp_oflow) << 32 |
5190                 le32_to_cpu(stat_info->rmac_icmp);
5191         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_tcp);
5192         tmp_stats[i++] = (u64)le32_to_cpu(stat_info->rmac_udp_oflow) << 32 |
5193                 le32_to_cpu(stat_info->rmac_udp);
5194         tmp_stats[i++] =
5195                 (u64)le32_to_cpu(stat_info->rmac_err_drp_udp_oflow) << 32 |
5196                 le32_to_cpu(stat_info->rmac_err_drp_udp);
5197         tmp_stats[i++] =
5198                 (u64)le32_to_cpu(stat_info->rmac_pause_cnt_oflow) << 32 |
5199                 le32_to_cpu(stat_info->rmac_pause_cnt);
5200         tmp_stats[i++] =
5201                 (u64)le32_to_cpu(stat_info->rmac_accepted_ip_oflow) << 32 |
5202                 le32_to_cpu(stat_info->rmac_accepted_ip);
5203         tmp_stats[i++] = le32_to_cpu(stat_info->rmac_err_tcp);
5204         tmp_stats[i++] = 0;
5205         tmp_stats[i++] = stat_info->sw_stat.single_ecc_errs;
5206         tmp_stats[i++] = stat_info->sw_stat.double_ecc_errs;
5207         tmp_stats[i++] = stat_info->sw_stat.clubbed_frms_cnt;
5208         tmp_stats[i++] = stat_info->sw_stat.sending_both;
5209         tmp_stats[i++] = stat_info->sw_stat.outof_sequence_pkts;
5210         tmp_stats[i++] = stat_info->sw_stat.flush_max_pkts;
5211         tmp = 0;
5212         if (stat_info->sw_stat.num_aggregations) {
5213                 tmp = stat_info->sw_stat.sum_avg_pkts_aggregated;
5214                 do_div(tmp, stat_info->sw_stat.num_aggregations);
5215         }
5216         tmp_stats[i++] = tmp;
5217 }
5218
5219 static int s2io_ethtool_get_regs_len(struct net_device *dev)
5220 {
5221         return (XENA_REG_SPACE);
5222 }
5223
5224
5225 static u32 s2io_ethtool_get_rx_csum(struct net_device * dev)
5226 {
5227         nic_t *sp = dev->priv;
5228
5229         return (sp->rx_csum);
5230 }
5231
5232 static int s2io_ethtool_set_rx_csum(struct net_device *dev, u32 data)
5233 {
5234         nic_t *sp = dev->priv;
5235
5236         if (data)
5237                 sp->rx_csum = 1;
5238         else
5239                 sp->rx_csum = 0;
5240
5241         return 0;
5242 }
5243
5244 static int s2io_get_eeprom_len(struct net_device *dev)
5245 {
5246         return (XENA_EEPROM_SPACE);
5247 }
5248
5249 static int s2io_ethtool_self_test_count(struct net_device *dev)
5250 {
5251         return (S2IO_TEST_LEN);
5252 }
5253
5254 static void s2io_ethtool_get_strings(struct net_device *dev,
5255                                      u32 stringset, u8 * data)
5256 {
5257         switch (stringset) {
5258         case ETH_SS_TEST:
5259                 memcpy(data, s2io_gstrings, S2IO_STRINGS_LEN);
5260                 break;
5261         case ETH_SS_STATS:
5262                 memcpy(data, &ethtool_stats_keys,
5263                        sizeof(ethtool_stats_keys));
5264         }
5265 }
5266 static int s2io_ethtool_get_stats_count(struct net_device *dev)
5267 {
5268         return (S2IO_STAT_LEN);
5269 }
5270
5271 static int s2io_ethtool_op_set_tx_csum(struct net_device *dev, u32 data)
5272 {
5273         if (data)
5274                 dev->features |= NETIF_F_IP_CSUM;
5275         else
5276                 dev->features &= ~NETIF_F_IP_CSUM;
5277
5278         return 0;
5279 }
5280
5281
5282 static struct ethtool_ops netdev_ethtool_ops = {
5283         .get_settings = s2io_ethtool_gset,
5284         .set_settings = s2io_ethtool_sset,
5285         .get_drvinfo = s2io_ethtool_gdrvinfo,
5286         .get_regs_len = s2io_ethtool_get_regs_len,
5287         .get_regs = s2io_ethtool_gregs,
5288         .get_link = ethtool_op_get_link,
5289         .get_eeprom_len = s2io_get_eeprom_len,
5290         .get_eeprom = s2io_ethtool_geeprom,
5291         .set_eeprom = s2io_ethtool_seeprom,
5292         .get_pauseparam = s2io_ethtool_getpause_data,
5293         .set_pauseparam = s2io_ethtool_setpause_data,
5294         .get_rx_csum = s2io_ethtool_get_rx_csum,
5295         .set_rx_csum = s2io_ethtool_set_rx_csum,
5296         .get_tx_csum = ethtool_op_get_tx_csum,
5297         .set_tx_csum = s2io_ethtool_op_set_tx_csum,
5298         .get_sg = ethtool_op_get_sg,
5299         .set_sg = ethtool_op_set_sg,
5300 #ifdef NETIF_F_TSO
5301         .get_tso = ethtool_op_get_tso,
5302         .set_tso = ethtool_op_set_tso,
5303 #endif
5304         .get_ufo = ethtool_op_get_ufo,
5305         .set_ufo = ethtool_op_set_ufo,
5306         .self_test_count = s2io_ethtool_self_test_count,
5307         .self_test = s2io_ethtool_test,
5308         .get_strings = s2io_ethtool_get_strings,
5309         .phys_id = s2io_ethtool_idnic,
5310         .get_stats_count = s2io_ethtool_get_stats_count,
5311         .get_ethtool_stats = s2io_get_ethtool_stats
5312 };
5313
5314 /**
5315  *  s2io_ioctl - Entry point for the Ioctl
5316  *  @dev :  Device pointer.
5317  *  @ifr :  An IOCTL specefic structure, that can contain a pointer to
5318  *  a proprietary structure used to pass information to the driver.
5319  *  @cmd :  This is used to distinguish between the different commands that
5320  *  can be passed to the IOCTL functions.
5321  *  Description:
5322  *  Currently there are no special functionality supported in IOCTL, hence
5323  *  function always return EOPNOTSUPPORTED
5324  */
5325
5326 static int s2io_ioctl(struct net_device *dev, struct ifreq *rq, int cmd)
5327 {
5328         return -EOPNOTSUPP;
5329 }
5330
5331 /**
5332  *  s2io_change_mtu - entry point to change MTU size for the device.
5333  *   @dev : device pointer.
5334  *   @new_mtu : the new MTU size for the device.
5335  *   Description: A driver entry point to change MTU size for the device.
5336  *   Before changing the MTU the device must be stopped.
5337  *  Return value:
5338  *   0 on success and an appropriate (-)ve integer as defined in errno.h
5339  *   file on failure.
5340  */
5341
5342 static int s2io_change_mtu(struct net_device *dev, int new_mtu)
5343 {
5344         nic_t *sp = dev->priv;
5345
5346         if ((new_mtu < MIN_MTU) || (new_mtu > S2IO_JUMBO_SIZE)) {
5347                 DBG_PRINT(ERR_DBG, "%s: MTU size is invalid.\n",
5348                           dev->name);
5349                 return -EPERM;
5350         }
5351
5352         dev->mtu = new_mtu;
5353         if (netif_running(dev)) {
5354                 s2io_card_down(sp);
5355                 netif_stop_queue(dev);
5356                 if (s2io_card_up(sp)) {
5357                         DBG_PRINT(ERR_DBG, "%s: Device bring up failed\n",
5358                                   __FUNCTION__);
5359                 }
5360                 if (netif_queue_stopped(dev))
5361                         netif_wake_queue(dev);
5362         } else { /* Device is down */
5363                 XENA_dev_config_t __iomem *bar0 = sp->bar0;
5364                 u64 val64 = new_mtu;
5365
5366                 writeq(vBIT(val64, 2, 14), &bar0->rmac_max_pyld_len);
5367         }
5368
5369         return 0;
5370 }
5371
5372 /**
5373  *  s2io_tasklet - Bottom half of the ISR.
5374  *  @dev_adr : address of the device structure in dma_addr_t format.
5375  *  Description:
5376  *  This is the tasklet or the bottom half of the ISR. This is
5377  *  an extension of the ISR which is scheduled by the scheduler to be run
5378  *  when the load on the CPU is low. All low priority tasks of the ISR can
5379  *  be pushed into the tasklet. For now the tasklet is used only to
5380  *  replenish the Rx buffers in the Rx buffer descriptors.
5381  *  Return value:
5382  *  void.
5383  */
5384
5385 static void s2io_tasklet(unsigned long dev_addr)
5386 {
5387         struct net_device *dev = (struct net_device *) dev_addr;
5388         nic_t *sp = dev->priv;
5389         int i, ret;
5390         mac_info_t *mac_control;
5391         struct config_param *config;
5392
5393         mac_control = &sp->mac_control;
5394         config = &sp->config;
5395
5396         if (!TASKLET_IN_USE) {
5397                 for (i = 0; i < config->rx_ring_num; i++) {
5398                         ret = fill_rx_buffers(sp, i);
5399                         if (ret == -ENOMEM) {
5400                                 DBG_PRINT(ERR_DBG, "%s: Out of ",
5401                                           dev->name);
5402                                 DBG_PRINT(ERR_DBG, "memory in tasklet\n");
5403                                 break;
5404                         } else if (ret == -EFILL) {
5405                                 DBG_PRINT(ERR_DBG,
5406                                           "%s: Rx Ring %d is full\n",
5407                                           dev->name, i);
5408                                 break;
5409                         }
5410                 }
5411                 clear_bit(0, (&sp->tasklet_status));
5412         }
5413 }
5414
5415 /**
5416  * s2io_set_link - Set the LInk status
5417  * @data: long pointer to device private structue
5418  * Description: Sets the link status for the adapter
5419  */
5420
5421 static void s2io_set_link(unsigned long data)
5422 {
5423         nic_t *nic = (nic_t *) data;
5424         struct net_device *dev = nic->dev;
5425         XENA_dev_config_t __iomem *bar0 = nic->bar0;
5426         register u64 val64;
5427         u16 subid;
5428
5429         if (test_and_set_bit(0, &(nic->link_state))) {
5430                 /* The card is being reset, no point doing anything */
5431                 return;
5432         }
5433
5434         subid = nic->pdev->subsystem_device;
5435         if (s2io_link_fault_indication(nic) == MAC_RMAC_ERR_TIMER) {
5436                 /*
5437                  * Allow a small delay for the NICs self initiated
5438                  * cleanup to complete.
5439                  */
5440                 msleep(100);
5441         }
5442
5443         val64 = readq(&bar0->adapter_status);
5444         if (verify_xena_quiescence(nic, val64, nic->device_enabled_once)) {
5445                 if (LINK_IS_UP(val64)) {
5446                         val64 = readq(&bar0->adapter_control);
5447                         val64 |= ADAPTER_CNTL_EN;
5448                         writeq(val64, &bar0->adapter_control);
5449                         if (CARDS_WITH_FAULTY_LINK_INDICATORS(nic->device_type,
5450                                                              subid)) {
5451                                 val64 = readq(&bar0->gpio_control);
5452                                 val64 |= GPIO_CTRL_GPIO_0;
5453                                 writeq(val64, &bar0->gpio_control);
5454                                 val64 = readq(&bar0->gpio_control);
5455                         } else {
5456                                 val64 |= ADAPTER_LED_ON;
5457                                 writeq(val64, &bar0->adapter_control);
5458                         }
5459                         if (s2io_link_fault_indication(nic) ==
5460                                                 MAC_RMAC_ERR_TIMER) {
5461                                 val64 = readq(&bar0->adapter_status);
5462                                 if (!LINK_IS_UP(val64)) {
5463                                         DBG_PRINT(ERR_DBG, "%s:", dev->name);
5464                                         DBG_PRINT(ERR_DBG, " Link down");
5465                                         DBG_PRINT(ERR_DBG, "after ");
5466                                         DBG_PRINT(ERR_DBG, "enabling ");
5467                                         DBG_PRINT(ERR_DBG, "device \n");
5468                                 }
5469                         }
5470                         if (nic->device_enabled_once == FALSE) {
5471                                 nic->device_enabled_once = TRUE;
5472                         }
5473                         s2io_link(nic, LINK_UP);
5474                 } else {
5475                         if (CARDS_WITH_FAULTY_LINK_INDICATORS(nic->device_type,
5476                                                               subid)) {
5477                                 val64 = readq(&bar0->gpio_control);
5478                                 val64 &= ~GPIO_CTRL_GPIO_0;
5479                                 writeq(val64, &bar0->gpio_control);
5480                                 val64 = readq(&bar0->gpio_control);
5481                         }
5482                         s2io_link(nic, LINK_DOWN);
5483                 }
5484         } else {                /* NIC is not Quiescent. */
5485                 DBG_PRINT(ERR_DBG, "%s: Error: ", dev->name);
5486                 DBG_PRINT(ERR_DBG, "device is not Quiescent\n");
5487                 netif_stop_queue(dev);
5488         }
5489         clear_bit(0, &(nic->link_state));
5490 }
5491
5492 static void s2io_card_down(nic_t * sp)
5493 {
5494         int cnt = 0;
5495         XENA_dev_config_t __iomem *bar0 = sp->bar0;
5496         unsigned long flags;
5497         register u64 val64 = 0;
5498
5499         del_timer_sync(&sp->alarm_timer);
5500         /* If s2io_set_link task is executing, wait till it completes. */
5501         while (test_and_set_bit(0, &(sp->link_state))) {
5502                 msleep(50);
5503         }
5504         atomic_set(&sp->card_state, CARD_DOWN);
5505
5506         /* disable Tx and Rx traffic on the NIC */
5507         stop_nic(sp);
5508
5509         /* Kill tasklet. */
5510         tasklet_kill(&sp->task);
5511
5512         /* Check if the device is Quiescent and then Reset the NIC */
5513         do {
5514                 val64 = readq(&bar0->adapter_status);
5515                 if (verify_xena_quiescence(sp, val64, sp->device_enabled_once)) {
5516                         break;
5517                 }
5518
5519                 msleep(50);
5520                 cnt++;
5521                 if (cnt == 10) {
5522                         DBG_PRINT(ERR_DBG,
5523                                   "s2io_close:Device not Quiescent ");
5524                         DBG_PRINT(ERR_DBG, "adaper status reads 0x%llx\n",
5525                                   (unsigned long long) val64);
5526                         break;
5527                 }
5528         } while (1);
5529         s2io_reset(sp);
5530
5531         /* Waiting till all Interrupt handlers are complete */
5532         cnt = 0;
5533         do {
5534                 msleep(10);
5535                 if (!atomic_read(&sp->isr_cnt))
5536                         break;
5537                 cnt++;
5538         } while(cnt < 5);
5539
5540         spin_lock_irqsave(&sp->tx_lock, flags);
5541         /* Free all Tx buffers */
5542         free_tx_buffers(sp);
5543         spin_unlock_irqrestore(&sp->tx_lock, flags);
5544
5545         /* Free all Rx buffers */
5546         spin_lock_irqsave(&sp->rx_lock, flags);
5547         free_rx_buffers(sp);
5548         spin_unlock_irqrestore(&sp->rx_lock, flags);
5549
5550         clear_bit(0, &(sp->link_state));
5551 }
5552
5553 static int s2io_card_up(nic_t * sp)
5554 {
5555         int i, ret = 0;
5556         mac_info_t *mac_control;
5557         struct config_param *config;
5558         struct net_device *dev = (struct net_device *) sp->dev;
5559
5560         /* Initialize the H/W I/O registers */
5561         if (init_nic(sp) != 0) {
5562                 DBG_PRINT(ERR_DBG, "%s: H/W initialization failed\n",
5563                           dev->name);
5564                 return -ENODEV;
5565         }
5566
5567         if (sp->intr_type == MSI)
5568                 ret = s2io_enable_msi(sp);
5569         else if (sp->intr_type == MSI_X)
5570                 ret = s2io_enable_msi_x(sp);
5571         if (ret) {
5572                 DBG_PRINT(ERR_DBG, "%s: Defaulting to INTA\n", dev->name);
5573                 sp->intr_type = INTA;
5574         }
5575
5576         /*
5577          * Initializing the Rx buffers. For now we are considering only 1
5578          * Rx ring and initializing buffers into 30 Rx blocks
5579          */
5580         mac_control = &sp->mac_control;
5581         config = &sp->config;
5582
5583         for (i = 0; i < config->rx_ring_num; i++) {
5584                 if ((ret = fill_rx_buffers(sp, i))) {
5585                         DBG_PRINT(ERR_DBG, "%s: Out of memory in Open\n",
5586                                   dev->name);
5587                         s2io_reset(sp);
5588                         free_rx_buffers(sp);
5589                         return -ENOMEM;
5590                 }
5591                 DBG_PRINT(INFO_DBG, "Buf in ring:%d is %d:\n", i,
5592                           atomic_read(&sp->rx_bufs_left[i]));
5593         }
5594
5595         /* Setting its receive mode */
5596         s2io_set_multicast(dev);
5597
5598         if (sp->lro) {
5599                 /* Initialize max aggregatable pkts based on MTU */
5600                 sp->lro_max_aggr_per_sess = ((1<<16) - 1) / dev->mtu;
5601                 /* Check if we can use(if specified) user provided value */
5602                 if (lro_max_pkts < sp->lro_max_aggr_per_sess)
5603                         sp->lro_max_aggr_per_sess = lro_max_pkts;
5604         }
5605
5606         /* Enable tasklet for the device */
5607         tasklet_init(&sp->task, s2io_tasklet, (unsigned long) dev);
5608
5609         /* Enable Rx Traffic and interrupts on the NIC */
5610         if (start_nic(sp)) {
5611                 DBG_PRINT(ERR_DBG, "%s: Starting NIC failed\n", dev->name);
5612                 tasklet_kill(&sp->task);
5613                 s2io_reset(sp);
5614                 free_irq(dev->irq, dev);
5615                 free_rx_buffers(sp);
5616                 return -ENODEV;
5617         }
5618
5619         S2IO_TIMER_CONF(sp->alarm_timer, s2io_alarm_handle, sp, (HZ/2));
5620
5621         atomic_set(&sp->card_state, CARD_UP);
5622         return 0;
5623 }
5624
5625 /**
5626  * s2io_restart_nic - Resets the NIC.
5627  * @data : long pointer to the device private structure
5628  * Description:
5629  * This function is scheduled to be run by the s2io_tx_watchdog
5630  * function after 0.5 secs to reset the NIC. The idea is to reduce
5631  * the run time of the watch dog routine which is run holding a
5632  * spin lock.
5633  */
5634
5635 static void s2io_restart_nic(unsigned long data)
5636 {
5637         struct net_device *dev = (struct net_device *) data;
5638         nic_t *sp = dev->priv;
5639
5640         s2io_card_down(sp);
5641         if (s2io_card_up(sp)) {
5642                 DBG_PRINT(ERR_DBG, "%s: Device bring up failed\n",
5643                           dev->name);
5644         }
5645         netif_wake_queue(dev);
5646         DBG_PRINT(ERR_DBG, "%s: was reset by Tx watchdog timer\n",
5647                   dev->name);
5648
5649 }
5650
5651 /**
5652  *  s2io_tx_watchdog - Watchdog for transmit side.
5653  *  @dev : Pointer to net device structure
5654  *  Description:
5655  *  This function is triggered if the Tx Queue is stopped
5656  *  for a pre-defined amount of time when the Interface is still up.
5657  *  If the Interface is jammed in such a situation, the hardware is
5658  *  reset (by s2io_close) and restarted again (by s2io_open) to
5659  *  overcome any problem that might have been caused in the hardware.
5660  *  Return value:
5661  *  void
5662  */
5663
5664 static void s2io_tx_watchdog(struct net_device *dev)
5665 {
5666         nic_t *sp = dev->priv;
5667
5668         if (netif_carrier_ok(dev)) {
5669                 schedule_work(&sp->rst_timer_task);
5670         }
5671 }
5672
5673 /**
5674  *   rx_osm_handler - To perform some OS related operations on SKB.
5675  *   @sp: private member of the device structure,pointer to s2io_nic structure.
5676  *   @skb : the socket buffer pointer.
5677  *   @len : length of the packet
5678  *   @cksum : FCS checksum of the frame.
5679  *   @ring_no : the ring from which this RxD was extracted.
5680  *   Description:
5681  *   This function is called by the Tx interrupt serivce routine to perform
5682  *   some OS related operations on the SKB before passing it to the upper
5683  *   layers. It mainly checks if the checksum is OK, if so adds it to the
5684  *   SKBs cksum variable, increments the Rx packet count and passes the SKB
5685  *   to the upper layer. If the checksum is wrong, it increments the Rx
5686  *   packet error count, frees the SKB and returns error.
5687  *   Return value:
5688  *   SUCCESS on success and -1 on failure.
5689  */
5690 static int rx_osm_handler(ring_info_t *ring_data, RxD_t * rxdp)
5691 {
5692         nic_t *sp = ring_data->nic;
5693         struct net_device *dev = (struct net_device *) sp->dev;
5694         struct sk_buff *skb = (struct sk_buff *)
5695                 ((unsigned long) rxdp->Host_Control);
5696         int ring_no = ring_data->ring_no;
5697         u16 l3_csum, l4_csum;
5698         lro_t *lro;
5699
5700         skb->dev = dev;
5701         if (rxdp->Control_1 & RXD_T_CODE) {
5702                 unsigned long long err = rxdp->Control_1 & RXD_T_CODE;
5703                 DBG_PRINT(ERR_DBG, "%s: Rx error Value: 0x%llx\n",
5704                           dev->name, err);
5705                 dev_kfree_skb(skb);
5706                 sp->stats.rx_crc_errors++;
5707                 atomic_dec(&sp->rx_bufs_left[ring_no]);
5708                 rxdp->Host_Control = 0;
5709                 return 0;
5710         }
5711
5712         /* Updating statistics */
5713         rxdp->Host_Control = 0;
5714         sp->rx_pkt_count++;
5715         sp->stats.rx_packets++;
5716         if (sp->rxd_mode == RXD_MODE_1) {
5717                 int len = RXD_GET_BUFFER0_SIZE_1(rxdp->Control_2);
5718
5719                 sp->stats.rx_bytes += len;
5720                 skb_put(skb, len);
5721
5722         } else if (sp->rxd_mode >= RXD_MODE_3A) {
5723                 int get_block = ring_data->rx_curr_get_info.block_index;
5724                 int get_off = ring_data->rx_curr_get_info.offset;
5725                 int buf0_len = RXD_GET_BUFFER0_SIZE_3(rxdp->Control_2);
5726                 int buf2_len = RXD_GET_BUFFER2_SIZE_3(rxdp->Control_2);
5727                 unsigned char *buff = skb_push(skb, buf0_len);
5728
5729                 buffAdd_t *ba = &ring_data->ba[get_block][get_off];
5730                 sp->stats.rx_bytes += buf0_len + buf2_len;
5731                 memcpy(buff, ba->ba_0, buf0_len);
5732
5733                 if (sp->rxd_mode == RXD_MODE_3A) {
5734                         int buf1_len = RXD_GET_BUFFER1_SIZE_3(rxdp->Control_2);
5735
5736                         skb_put(skb, buf1_len);
5737                         skb->len += buf2_len;
5738                         skb->data_len += buf2_len;
5739                         skb->truesize += buf2_len;
5740                         skb_put(skb_shinfo(skb)->frag_list, buf2_len);
5741                         sp->stats.rx_bytes += buf1_len;
5742
5743                 } else
5744                         skb_put(skb, buf2_len);
5745         }
5746
5747         if ((rxdp->Control_1 & TCP_OR_UDP_FRAME) && ((!sp->lro) ||
5748             (sp->lro && (!(rxdp->Control_1 & RXD_FRAME_IP_FRAG)))) &&
5749             (sp->rx_csum)) {
5750                 l3_csum = RXD_GET_L3_CKSUM(rxdp->Control_1);
5751                 l4_csum = RXD_GET_L4_CKSUM(rxdp->Control_1);
5752                 if ((l3_csum == L3_CKSUM_OK) && (l4_csum == L4_CKSUM_OK)) {
5753                         /*
5754                          * NIC verifies if the Checksum of the received
5755                          * frame is Ok or not and accordingly returns
5756                          * a flag in the RxD.
5757                          */
5758                         skb->ip_summed = CHECKSUM_UNNECESSARY;
5759                         if (sp->lro) {
5760                                 u32 tcp_len;
5761                                 u8 *tcp;
5762                                 int ret = 0;
5763
5764                                 ret = s2io_club_tcp_session(skb->data, &tcp,
5765                                                 &tcp_len, &lro, rxdp, sp);
5766                                 switch (ret) {
5767                                         case 3: /* Begin anew */
5768                                                 lro->parent = skb;
5769                                                 goto aggregate;
5770                                         case 1: /* Aggregate */
5771                                         {
5772                                                 lro_append_pkt(sp, lro,
5773                                                         skb, tcp_len);
5774                                                 goto aggregate;
5775                                         }
5776                                         case 4: /* Flush session */
5777                                         {
5778                                                 lro_append_pkt(sp, lro,
5779                                                         skb, tcp_len);
5780                                                 queue_rx_frame(lro->parent);
5781                                                 clear_lro_session(lro);
5782                                                 sp->mac_control.stats_info->
5783                                                     sw_stat.flush_max_pkts++;
5784                                                 goto aggregate;
5785                                         }
5786                                         case 2: /* Flush both */
5787                                                 lro->parent->data_len =
5788                                                         lro->frags_len;
5789                                                 sp->mac_control.stats_info->
5790                                                      sw_stat.sending_both++;
5791                                                 queue_rx_frame(lro->parent);
5792                                                 clear_lro_session(lro);
5793                                                 goto send_up;
5794                                         case 0: /* sessions exceeded */
5795                                         case 5: /*
5796                                                  * First pkt in session not
5797                                                  * L3/L4 aggregatable
5798                                                  */
5799                                                 break;
5800                                         default:
5801                                                 DBG_PRINT(ERR_DBG,
5802                                                         "%s: Samadhana!!\n",
5803                                                          __FUNCTION__);
5804                                                 BUG();
5805                                 }
5806                         }
5807                 } else {
5808                         /*
5809                          * Packet with erroneous checksum, let the
5810                          * upper layers deal with it.
5811                          */
5812                         skb->ip_summed = CHECKSUM_NONE;
5813                 }
5814         } else {
5815                 skb->ip_summed = CHECKSUM_NONE;
5816         }
5817
5818         if (!sp->lro) {
5819                 skb->protocol = eth_type_trans(skb, dev);
5820 #ifdef CONFIG_S2IO_NAPI
5821                 if (sp->vlgrp && RXD_GET_VLAN_TAG(rxdp->Control_2)) {
5822                         /* Queueing the vlan frame to the upper layer */
5823                         vlan_hwaccel_receive_skb(skb, sp->vlgrp,
5824                                 RXD_GET_VLAN_TAG(rxdp->Control_2));
5825                 } else {
5826                         netif_receive_skb(skb);
5827                 }
5828 #else
5829                 if (sp->vlgrp && RXD_GET_VLAN_TAG(rxdp->Control_2)) {
5830                         /* Queueing the vlan frame to the upper layer */
5831                         vlan_hwaccel_rx(skb, sp->vlgrp,
5832                                 RXD_GET_VLAN_TAG(rxdp->Control_2));
5833                 } else {
5834                         netif_rx(skb);
5835                 }
5836 #endif
5837         } else {
5838 send_up:
5839                 queue_rx_frame(skb);
5840         }
5841         dev->last_rx = jiffies;
5842 aggregate:
5843         atomic_dec(&sp->rx_bufs_left[ring_no]);
5844         return SUCCESS;
5845 }
5846
5847 /**
5848  *  s2io_link - stops/starts the Tx queue.
5849  *  @sp : private member of the device structure, which is a pointer to the
5850  *  s2io_nic structure.
5851  *  @link : inidicates whether link is UP/DOWN.
5852  *  Description:
5853  *  This function stops/starts the Tx queue depending on whether the link
5854  *  status of the NIC is is down or up. This is called by the Alarm
5855  *  interrupt handler whenever a link change interrupt comes up.
5856  *  Return value:
5857  *  void.
5858  */
5859
5860 static void s2io_link(nic_t * sp, int link)
5861 {
5862         struct net_device *dev = (struct net_device *) sp->dev;
5863
5864         if (link != sp->last_link_state) {
5865                 if (link == LINK_DOWN) {
5866                         DBG_PRINT(ERR_DBG, "%s: Link down\n", dev->name);
5867                         netif_carrier_off(dev);
5868                 } else {
5869                         DBG_PRINT(ERR_DBG, "%s: Link Up\n", dev->name);
5870                         netif_carrier_on(dev);
5871                 }
5872         }
5873         sp->last_link_state = link;
5874 }
5875
5876 /**
5877  *  get_xena_rev_id - to identify revision ID of xena.
5878  *  @pdev : PCI Dev structure
5879  *  Description:
5880  *  Function to identify the Revision ID of xena.
5881  *  Return value:
5882  *  returns the revision ID of the device.
5883  */
5884
5885 static int get_xena_rev_id(struct pci_dev *pdev)
5886 {
5887         u8 id = 0;
5888         int ret;
5889         ret = pci_read_config_byte(pdev, PCI_REVISION_ID, (u8 *) & id);
5890         return id;
5891 }
5892
5893 /**
5894  *  s2io_init_pci -Initialization of PCI and PCI-X configuration registers .
5895  *  @sp : private member of the device structure, which is a pointer to the
5896  *  s2io_nic structure.
5897  *  Description:
5898  *  This function initializes a few of the PCI and PCI-X configuration registers
5899  *  with recommended values.
5900  *  Return value:
5901  *  void
5902  */
5903
5904 static void s2io_init_pci(nic_t * sp)
5905 {
5906         u16 pci_cmd = 0, pcix_cmd = 0;
5907
5908         /* Enable Data Parity Error Recovery in PCI-X command register. */
5909         pci_read_config_word(sp->pdev, PCIX_COMMAND_REGISTER,
5910                              &(pcix_cmd));
5911         pci_write_config_word(sp->pdev, PCIX_COMMAND_REGISTER,
5912                               (pcix_cmd | 1));
5913         pci_read_config_word(sp->pdev, PCIX_COMMAND_REGISTER,
5914                              &(pcix_cmd));
5915
5916         /* Set the PErr Response bit in PCI command register. */
5917         pci_read_config_word(sp->pdev, PCI_COMMAND, &pci_cmd);
5918         pci_write_config_word(sp->pdev, PCI_COMMAND,
5919                               (pci_cmd | PCI_COMMAND_PARITY));
5920         pci_read_config_word(sp->pdev, PCI_COMMAND, &pci_cmd);
5921
5922         /* Forcibly disabling relaxed ordering capability of the card. */
5923         pcix_cmd &= 0xfffd;
5924         pci_write_config_word(sp->pdev, PCIX_COMMAND_REGISTER,
5925                               pcix_cmd);
5926         pci_read_config_word(sp->pdev, PCIX_COMMAND_REGISTER,
5927                              &(pcix_cmd));
5928 }
5929
5930 MODULE_AUTHOR("Raghavendra Koushik <raghavendra.koushik@neterion.com>");
5931 MODULE_LICENSE("GPL");
5932 MODULE_VERSION(DRV_VERSION);
5933
5934 module_param(tx_fifo_num, int, 0);
5935 module_param(rx_ring_num, int, 0);
5936 module_param(rx_ring_mode, int, 0);
5937 module_param_array(tx_fifo_len, uint, NULL, 0);
5938 module_param_array(rx_ring_sz, uint, NULL, 0);
5939 module_param_array(rts_frm_len, uint, NULL, 0);
5940 module_param(use_continuous_tx_intrs, int, 1);
5941 module_param(rmac_pause_time, int, 0);
5942 module_param(mc_pause_threshold_q0q3, int, 0);
5943 module_param(mc_pause_threshold_q4q7, int, 0);
5944 module_param(shared_splits, int, 0);
5945 module_param(tmac_util_period, int, 0);
5946 module_param(rmac_util_period, int, 0);
5947 module_param(bimodal, bool, 0);
5948 module_param(l3l4hdr_size, int , 0);
5949 #ifndef CONFIG_S2IO_NAPI
5950 module_param(indicate_max_pkts, int, 0);
5951 #endif
5952 module_param(rxsync_frequency, int, 0);
5953 module_param(intr_type, int, 0);
5954 module_param(lro, int, 0);
5955 module_param(lro_max_pkts, int, 0);
5956
5957 /**
5958  *  s2io_init_nic - Initialization of the adapter .
5959  *  @pdev : structure containing the PCI related information of the device.
5960  *  @pre: List of PCI devices supported by the driver listed in s2io_tbl.
5961  *  Description:
5962  *  The function initializes an adapter identified by the pci_dec structure.
5963  *  All OS related initialization including memory and device structure and
5964  *  initlaization of the device private variable is done. Also the swapper
5965  *  control register is initialized to enable read and write into the I/O
5966  *  registers of the device.
5967  *  Return value:
5968  *  returns 0 on success and negative on failure.
5969  */
5970
5971 static int __devinit
5972 s2io_init_nic(struct pci_dev *pdev, const struct pci_device_id *pre)
5973 {
5974         nic_t *sp;
5975         struct net_device *dev;
5976         int i, j, ret;
5977         int dma_flag = FALSE;
5978         u32 mac_up, mac_down;
5979         u64 val64 = 0, tmp64 = 0;
5980         XENA_dev_config_t __iomem *bar0 = NULL;
5981         u16 subid;
5982         mac_info_t *mac_control;
5983         struct config_param *config;
5984         int mode;
5985         u8 dev_intr_type = intr_type;
5986
5987 #ifdef CONFIG_S2IO_NAPI
5988         if (dev_intr_type != INTA) {
5989                 DBG_PRINT(ERR_DBG, "NAPI cannot be enabled when MSI/MSI-X \
5990 is enabled. Defaulting to INTA\n");
5991                 dev_intr_type = INTA;
5992         }
5993         else
5994                 DBG_PRINT(ERR_DBG, "NAPI support has been enabled\n");
5995 #endif
5996
5997         if ((ret = pci_enable_device(pdev))) {
5998                 DBG_PRINT(ERR_DBG,
5999                           "s2io_init_nic: pci_enable_device failed\n");
6000                 return ret;
6001         }
6002
6003         if (!pci_set_dma_mask(pdev, DMA_64BIT_MASK)) {
6004                 DBG_PRINT(INIT_DBG, "s2io_init_nic: Using 64bit DMA\n");
6005                 dma_flag = TRUE;
6006                 if (pci_set_consistent_dma_mask
6007                     (pdev, DMA_64BIT_MASK)) {
6008                         DBG_PRINT(ERR_DBG,
6009                                   "Unable to obtain 64bit DMA for \
6010                                         consistent allocations\n");
6011                         pci_disable_device(pdev);
6012                         return -ENOMEM;
6013                 }
6014         } else if (!pci_set_dma_mask(pdev, DMA_32BIT_MASK)) {
6015                 DBG_PRINT(INIT_DBG, "s2io_init_nic: Using 32bit DMA\n");
6016         } else {
6017                 pci_disable_device(pdev);
6018                 return -ENOMEM;
6019         }
6020
6021         if ((dev_intr_type == MSI_X) &&
6022                         ((pdev->device != PCI_DEVICE_ID_HERC_WIN) &&
6023                         (pdev->device != PCI_DEVICE_ID_HERC_UNI))) {
6024                 DBG_PRINT(ERR_DBG, "Xframe I does not support MSI_X. \
6025 Defaulting to INTA\n");
6026                 dev_intr_type = INTA;
6027         }
6028         if (dev_intr_type != MSI_X) {
6029                 if (pci_request_regions(pdev, s2io_driver_name)) {
6030                         DBG_PRINT(ERR_DBG, "Request Regions failed\n"),
6031                             pci_disable_device(pdev);
6032                         return -ENODEV;
6033                 }
6034         }
6035         else {
6036                 if (!(request_mem_region(pci_resource_start(pdev, 0),
6037                          pci_resource_len(pdev, 0), s2io_driver_name))) {
6038                         DBG_PRINT(ERR_DBG, "bar0 Request Regions failed\n");
6039                         pci_disable_device(pdev);
6040                         return -ENODEV;
6041                 }
6042                 if (!(request_mem_region(pci_resource_start(pdev, 2),
6043                          pci_resource_len(pdev, 2), s2io_driver_name))) {
6044                         DBG_PRINT(ERR_DBG, "bar1 Request Regions failed\n");
6045                         release_mem_region(pci_resource_start(pdev, 0),
6046                                    pci_resource_len(pdev, 0));
6047                         pci_disable_device(pdev);
6048                         return -ENODEV;
6049                 }
6050         }
6051
6052         dev = alloc_etherdev(sizeof(nic_t));
6053         if (dev == NULL) {
6054                 DBG_PRINT(ERR_DBG, "Device allocation failed\n");
6055                 pci_disable_device(pdev);
6056                 pci_release_regions(pdev);
6057                 return -ENODEV;
6058         }
6059
6060         pci_set_master(pdev);
6061         pci_set_drvdata(pdev, dev);
6062         SET_MODULE_OWNER(dev);
6063         SET_NETDEV_DEV(dev, &pdev->dev);
6064
6065         /*  Private member variable initialized to s2io NIC structure */
6066         sp = dev->priv;
6067         memset(sp, 0, sizeof(nic_t));
6068         sp->dev = dev;
6069         sp->pdev = pdev;
6070         sp->high_dma_flag = dma_flag;
6071         sp->device_enabled_once = FALSE;
6072         if (rx_ring_mode == 1)
6073                 sp->rxd_mode = RXD_MODE_1;
6074         if (rx_ring_mode == 2)
6075                 sp->rxd_mode = RXD_MODE_3B;
6076         if (rx_ring_mode == 3)
6077                 sp->rxd_mode = RXD_MODE_3A;
6078
6079         sp->intr_type = dev_intr_type;
6080
6081         if ((pdev->device == PCI_DEVICE_ID_HERC_WIN) ||
6082                 (pdev->device == PCI_DEVICE_ID_HERC_UNI))
6083                 sp->device_type = XFRAME_II_DEVICE;
6084         else
6085                 sp->device_type = XFRAME_I_DEVICE;
6086
6087         sp->lro = lro;
6088
6089         /* Initialize some PCI/PCI-X fields of the NIC. */
6090         s2io_init_pci(sp);
6091
6092         /*
6093          * Setting the device configuration parameters.
6094          * Most of these parameters can be specified by the user during
6095          * module insertion as they are module loadable parameters. If
6096          * these parameters are not not specified during load time, they
6097          * are initialized with default values.
6098          */
6099         mac_control = &sp->mac_control;
6100         config = &sp->config;
6101
6102         /* Tx side parameters. */
6103         if (tx_fifo_len[0] == 0)
6104                 tx_fifo_len[0] = DEFAULT_FIFO_LEN; /* Default value. */
6105         config->tx_fifo_num = tx_fifo_num;
6106         for (i = 0; i < MAX_TX_FIFOS; i++) {
6107                 config->tx_cfg[i].fifo_len = tx_fifo_len[i];
6108                 config->tx_cfg[i].fifo_priority = i;
6109         }
6110
6111         /* mapping the QoS priority to the configured fifos */
6112         for (i = 0; i < MAX_TX_FIFOS; i++)
6113                 config->fifo_mapping[i] = fifo_map[config->tx_fifo_num][i];
6114
6115         config->tx_intr_type = TXD_INT_TYPE_UTILZ;
6116         for (i = 0; i < config->tx_fifo_num; i++) {
6117                 config->tx_cfg[i].f_no_snoop =
6118                     (NO_SNOOP_TXD | NO_SNOOP_TXD_BUFFER);
6119                 if (config->tx_cfg[i].fifo_len < 65) {
6120                         config->tx_intr_type = TXD_INT_TYPE_PER_LIST;
6121                         break;
6122                 }
6123         }
6124         /* + 2 because one Txd for skb->data and one Txd for UFO */
6125         config->max_txds = MAX_SKB_FRAGS + 2;
6126
6127         /* Rx side parameters. */
6128         if (rx_ring_sz[0] == 0)
6129                 rx_ring_sz[0] = SMALL_BLK_CNT; /* Default value. */
6130         config->rx_ring_num = rx_ring_num;
6131         for (i = 0; i < MAX_RX_RINGS; i++) {
6132                 config->rx_cfg[i].num_rxd = rx_ring_sz[i] *
6133                     (rxd_count[sp->rxd_mode] + 1);
6134                 config->rx_cfg[i].ring_priority = i;
6135         }
6136
6137         for (i = 0; i < rx_ring_num; i++) {
6138                 config->rx_cfg[i].ring_org = RING_ORG_BUFF1;
6139                 config->rx_cfg[i].f_no_snoop =
6140                     (NO_SNOOP_RXD | NO_SNOOP_RXD_BUFFER);
6141         }
6142
6143         /*  Setting Mac Control parameters */
6144         mac_control->rmac_pause_time = rmac_pause_time;
6145         mac_control->mc_pause_threshold_q0q3 = mc_pause_threshold_q0q3;
6146         mac_control->mc_pause_threshold_q4q7 = mc_pause_threshold_q4q7;
6147
6148
6149         /* Initialize Ring buffer parameters. */
6150         for (i = 0; i < config->rx_ring_num; i++)
6151                 atomic_set(&sp->rx_bufs_left[i], 0);
6152
6153         /* Initialize the number of ISRs currently running */
6154         atomic_set(&sp->isr_cnt, 0);
6155
6156         /*  initialize the shared memory used by the NIC and the host */
6157         if (init_shared_mem(sp)) {
6158                 DBG_PRINT(ERR_DBG, "%s: Memory allocation failed\n",
6159                           __FUNCTION__);
6160                 ret = -ENOMEM;
6161                 goto mem_alloc_failed;
6162         }
6163
6164         sp->bar0 = ioremap(pci_resource_start(pdev, 0),
6165                                      pci_resource_len(pdev, 0));
6166         if (!sp->bar0) {
6167                 DBG_PRINT(ERR_DBG, "%s: S2IO: cannot remap io mem1\n",
6168                           dev->name);
6169                 ret = -ENOMEM;
6170                 goto bar0_remap_failed;
6171         }
6172
6173         sp->bar1 = ioremap(pci_resource_start(pdev, 2),
6174                                      pci_resource_len(pdev, 2));
6175         if (!sp->bar1) {
6176                 DBG_PRINT(ERR_DBG, "%s: S2IO: cannot remap io mem2\n",
6177                           dev->name);
6178                 ret = -ENOMEM;
6179                 goto bar1_remap_failed;
6180         }
6181
6182         dev->irq = pdev->irq;
6183         dev->base_addr = (unsigned long) sp->bar0;
6184
6185         /* Initializing the BAR1 address as the start of the FIFO pointer. */
6186         for (j = 0; j < MAX_TX_FIFOS; j++) {
6187                 mac_control->tx_FIFO_start[j] = (TxFIFO_element_t __iomem *)
6188                     (sp->bar1 + (j * 0x00020000));
6189         }
6190
6191         /*  Driver entry points */
6192         dev->open = &s2io_open;
6193         dev->stop = &s2io_close;
6194         dev->hard_start_xmit = &s2io_xmit;
6195         dev->get_stats = &s2io_get_stats;
6196         dev->set_multicast_list = &s2io_set_multicast;
6197         dev->do_ioctl = &s2io_ioctl;
6198         dev->change_mtu = &s2io_change_mtu;
6199         SET_ETHTOOL_OPS(dev, &netdev_ethtool_ops);
6200         dev->features |= NETIF_F_HW_VLAN_TX | NETIF_F_HW_VLAN_RX;
6201         dev->vlan_rx_register = s2io_vlan_rx_register;
6202         dev->vlan_rx_kill_vid = (void *)s2io_vlan_rx_kill_vid;
6203
6204         /*
6205          * will use eth_mac_addr() for  dev->set_mac_address
6206          * mac address will be set every time dev->open() is called
6207          */
6208 #if defined(CONFIG_S2IO_NAPI)
6209         dev->poll = s2io_poll;
6210         dev->weight = 32;
6211 #endif
6212
6213         dev->features |= NETIF_F_SG | NETIF_F_IP_CSUM;
6214         if (sp->high_dma_flag == TRUE)
6215                 dev->features |= NETIF_F_HIGHDMA;
6216 #ifdef NETIF_F_TSO
6217         dev->features |= NETIF_F_TSO;
6218 #endif
6219         if (sp->device_type & XFRAME_II_DEVICE) {
6220                 dev->features |= NETIF_F_UFO;
6221                 dev->features |= NETIF_F_HW_CSUM;
6222         }
6223
6224         dev->tx_timeout = &s2io_tx_watchdog;
6225         dev->watchdog_timeo = WATCH_DOG_TIMEOUT;
6226         INIT_WORK(&sp->rst_timer_task,
6227                   (void (*)(void *)) s2io_restart_nic, dev);
6228         INIT_WORK(&sp->set_link_task,
6229                   (void (*)(void *)) s2io_set_link, sp);
6230
6231         pci_save_state(sp->pdev);
6232
6233         /* Setting swapper control on the NIC, for proper reset operation */
6234         if (s2io_set_swapper(sp)) {
6235                 DBG_PRINT(ERR_DBG, "%s:swapper settings are wrong\n",
6236                           dev->name);
6237                 ret = -EAGAIN;
6238                 goto set_swap_failed;
6239         }
6240
6241         /* Verify if the Herc works on the slot its placed into */
6242         if (sp->device_type & XFRAME_II_DEVICE) {
6243                 mode = s2io_verify_pci_mode(sp);
6244                 if (mode < 0) {
6245                         DBG_PRINT(ERR_DBG, "%s: ", __FUNCTION__);
6246                         DBG_PRINT(ERR_DBG, " Unsupported PCI bus mode\n");
6247                         ret = -EBADSLT;
6248                         goto set_swap_failed;
6249                 }
6250         }
6251
6252         /* Not needed for Herc */
6253         if (sp->device_type & XFRAME_I_DEVICE) {
6254                 /*
6255                  * Fix for all "FFs" MAC address problems observed on
6256                  * Alpha platforms
6257                  */
6258                 fix_mac_address(sp);
6259                 s2io_reset(sp);
6260         }
6261
6262         /*
6263          * MAC address initialization.
6264          * For now only one mac address will be read and used.
6265          */
6266         bar0 = sp->bar0;
6267         val64 = RMAC_ADDR_CMD_MEM_RD | RMAC_ADDR_CMD_MEM_STROBE_NEW_CMD |
6268             RMAC_ADDR_CMD_MEM_OFFSET(0 + MAC_MAC_ADDR_START_OFFSET);
6269         writeq(val64, &bar0->rmac_addr_cmd_mem);
6270         wait_for_cmd_complete(sp);
6271
6272         tmp64 = readq(&bar0->rmac_addr_data0_mem);
6273         mac_down = (u32) tmp64;
6274         mac_up = (u32) (tmp64 >> 32);
6275
6276         memset(sp->def_mac_addr[0].mac_addr, 0, sizeof(ETH_ALEN));
6277
6278         sp->def_mac_addr[0].mac_addr[3] = (u8) (mac_up);
6279         sp->def_mac_addr[0].mac_addr[2] = (u8) (mac_up >> 8);
6280         sp->def_mac_addr[0].mac_addr[1] = (u8) (mac_up >> 16);
6281         sp->def_mac_addr[0].mac_addr[0] = (u8) (mac_up >> 24);
6282         sp->def_mac_addr[0].mac_addr[5] = (u8) (mac_down >> 16);
6283         sp->def_mac_addr[0].mac_addr[4] = (u8) (mac_down >> 24);
6284
6285         /*  Set the factory defined MAC address initially   */
6286         dev->addr_len = ETH_ALEN;
6287         memcpy(dev->dev_addr, sp->def_mac_addr, ETH_ALEN);
6288
6289         /*
6290          * Initialize the tasklet status and link state flags
6291          * and the card state parameter
6292          */
6293         atomic_set(&(sp->card_state), 0);
6294         sp->tasklet_status = 0;
6295         sp->link_state = 0;
6296
6297         /* Initialize spinlocks */
6298         spin_lock_init(&sp->tx_lock);
6299 #ifndef CONFIG_S2IO_NAPI
6300         spin_lock_init(&sp->put_lock);
6301 #endif
6302         spin_lock_init(&sp->rx_lock);
6303
6304         /*
6305          * SXE-002: Configure link and activity LED to init state
6306          * on driver load.
6307          */
6308         subid = sp->pdev->subsystem_device;
6309         if ((subid & 0xFF) >= 0x07) {
6310                 val64 = readq(&bar0->gpio_control);
6311                 val64 |= 0x0000800000000000ULL;
6312                 writeq(val64, &bar0->gpio_control);
6313                 val64 = 0x0411040400000000ULL;
6314                 writeq(val64, (void __iomem *) bar0 + 0x2700);
6315                 val64 = readq(&bar0->gpio_control);
6316         }
6317
6318         sp->rx_csum = 1;        /* Rx chksum verify enabled by default */
6319
6320         if (register_netdev(dev)) {
6321                 DBG_PRINT(ERR_DBG, "Device registration failed\n");
6322                 ret = -ENODEV;
6323                 goto register_failed;
6324         }
6325
6326         if (sp->device_type & XFRAME_II_DEVICE) {
6327                 DBG_PRINT(ERR_DBG, "%s: Neterion Xframe II 10GbE adapter ",
6328                           dev->name);
6329                 DBG_PRINT(ERR_DBG, "(rev %d), Version %s",
6330                                 get_xena_rev_id(sp->pdev),
6331                                 s2io_driver_version);
6332                 switch(sp->intr_type) {
6333                         case INTA:
6334                                 DBG_PRINT(ERR_DBG, ", Intr type INTA");
6335                                 break;
6336                         case MSI:
6337                                 DBG_PRINT(ERR_DBG, ", Intr type MSI");
6338                                 break;
6339                         case MSI_X:
6340                                 DBG_PRINT(ERR_DBG, ", Intr type MSI-X");
6341                                 break;
6342                 }
6343
6344                 DBG_PRINT(ERR_DBG, "\nCopyright(c) 2002-2005 Neterion Inc.\n");
6345                 DBG_PRINT(ERR_DBG, "MAC ADDR: %02x:%02x:%02x:%02x:%02x:%02x\n",
6346                           sp->def_mac_addr[0].mac_addr[0],
6347                           sp->def_mac_addr[0].mac_addr[1],
6348                           sp->def_mac_addr[0].mac_addr[2],
6349                           sp->def_mac_addr[0].mac_addr[3],
6350                           sp->def_mac_addr[0].mac_addr[4],
6351                           sp->def_mac_addr[0].mac_addr[5]);
6352                 mode = s2io_print_pci_mode(sp);
6353                 if (mode < 0) {
6354                         DBG_PRINT(ERR_DBG, " Unsupported PCI bus mode ");
6355                         ret = -EBADSLT;
6356                         goto set_swap_failed;
6357                 }
6358         } else {
6359                 DBG_PRINT(ERR_DBG, "%s: Neterion Xframe I 10GbE adapter ",
6360                           dev->name);
6361                 DBG_PRINT(ERR_DBG, "(rev %d), Version %s",
6362                                         get_xena_rev_id(sp->pdev),
6363                                         s2io_driver_version);
6364                 switch(sp->intr_type) {
6365                         case INTA:
6366                                 DBG_PRINT(ERR_DBG, ", Intr type INTA");
6367                                 break;
6368                         case MSI:
6369                                 DBG_PRINT(ERR_DBG, ", Intr type MSI");
6370                                 break;
6371                         case MSI_X:
6372                                 DBG_PRINT(ERR_DBG, ", Intr type MSI-X");
6373                                 break;
6374                 }
6375                 DBG_PRINT(ERR_DBG, "\nCopyright(c) 2002-2005 Neterion Inc.\n");
6376                 DBG_PRINT(ERR_DBG, "MAC ADDR: %02x:%02x:%02x:%02x:%02x:%02x\n",
6377                           sp->def_mac_addr[0].mac_addr[0],
6378                           sp->def_mac_addr[0].mac_addr[1],
6379                           sp->def_mac_addr[0].mac_addr[2],
6380                           sp->def_mac_addr[0].mac_addr[3],
6381                           sp->def_mac_addr[0].mac_addr[4],
6382                           sp->def_mac_addr[0].mac_addr[5]);
6383         }
6384         if (sp->rxd_mode == RXD_MODE_3B)
6385                 DBG_PRINT(ERR_DBG, "%s: 2-Buffer mode support has been "
6386                           "enabled\n",dev->name);
6387         if (sp->rxd_mode == RXD_MODE_3A)
6388                 DBG_PRINT(ERR_DBG, "%s: 3-Buffer mode support has been "
6389                           "enabled\n",dev->name);
6390
6391         if (sp->lro)
6392                 DBG_PRINT(ERR_DBG, "%s: Large receive offload enabled\n",
6393                         dev->name);
6394
6395         /* Initialize device name */
6396         strcpy(sp->name, dev->name);
6397         if (sp->device_type & XFRAME_II_DEVICE)
6398                 strcat(sp->name, ": Neterion Xframe II 10GbE adapter");
6399         else
6400                 strcat(sp->name, ": Neterion Xframe I 10GbE adapter");
6401
6402         /* Initialize bimodal Interrupts */
6403         sp->config.bimodal = bimodal;
6404         if (!(sp->device_type & XFRAME_II_DEVICE) && bimodal) {
6405                 sp->config.bimodal = 0;
6406                 DBG_PRINT(ERR_DBG,"%s:Bimodal intr not supported by Xframe I\n",
6407                         dev->name);
6408         }
6409
6410         /*
6411          * Make Link state as off at this point, when the Link change
6412          * interrupt comes the state will be automatically changed to
6413          * the right state.
6414          */
6415         netif_carrier_off(dev);
6416
6417         return 0;
6418
6419       register_failed:
6420       set_swap_failed:
6421         iounmap(sp->bar1);
6422       bar1_remap_failed:
6423         iounmap(sp->bar0);
6424       bar0_remap_failed:
6425       mem_alloc_failed:
6426         free_shared_mem(sp);
6427         pci_disable_device(pdev);
6428         if (dev_intr_type != MSI_X)
6429                 pci_release_regions(pdev);
6430         else {
6431                 release_mem_region(pci_resource_start(pdev, 0),
6432                         pci_resource_len(pdev, 0));
6433                 release_mem_region(pci_resource_start(pdev, 2),
6434                         pci_resource_len(pdev, 2));
6435         }
6436         pci_set_drvdata(pdev, NULL);
6437         free_netdev(dev);
6438
6439         return ret;
6440 }
6441
6442 /**
6443  * s2io_rem_nic - Free the PCI device
6444  * @pdev: structure containing the PCI related information of the device.
6445  * Description: This function is called by the Pci subsystem to release a
6446  * PCI device and free up all resource held up by the device. This could
6447  * be in response to a Hot plug event or when the driver is to be removed
6448  * from memory.
6449  */
6450
6451 static void __devexit s2io_rem_nic(struct pci_dev *pdev)
6452 {
6453         struct net_device *dev =
6454             (struct net_device *) pci_get_drvdata(pdev);
6455         nic_t *sp;
6456
6457         if (dev == NULL) {
6458                 DBG_PRINT(ERR_DBG, "Driver Data is NULL!!\n");
6459                 return;
6460         }
6461
6462         sp = dev->priv;
6463         unregister_netdev(dev);
6464
6465         free_shared_mem(sp);
6466         iounmap(sp->bar0);
6467         iounmap(sp->bar1);
6468         pci_disable_device(pdev);
6469         if (sp->intr_type != MSI_X)
6470                 pci_release_regions(pdev);
6471         else {
6472                 release_mem_region(pci_resource_start(pdev, 0),
6473                         pci_resource_len(pdev, 0));
6474                 release_mem_region(pci_resource_start(pdev, 2),
6475                         pci_resource_len(pdev, 2));
6476         }
6477         pci_set_drvdata(pdev, NULL);
6478         free_netdev(dev);
6479 }
6480
6481 /**
6482  * s2io_starter - Entry point for the driver
6483  * Description: This function is the entry point for the driver. It verifies
6484  * the module loadable parameters and initializes PCI configuration space.
6485  */
6486
6487 int __init s2io_starter(void)
6488 {
6489         return pci_module_init(&s2io_driver);
6490 }
6491
6492 /**
6493  * s2io_closer - Cleanup routine for the driver
6494  * Description: This function is the cleanup routine for the driver. It unregist * ers the driver.
6495  */
6496
6497 static void s2io_closer(void)
6498 {
6499         pci_unregister_driver(&s2io_driver);
6500         DBG_PRINT(INIT_DBG, "cleanup done\n");
6501 }
6502
6503 module_init(s2io_starter);
6504 module_exit(s2io_closer);
6505
6506 static int check_L2_lro_capable(u8 *buffer, struct iphdr **ip,
6507                 struct tcphdr **tcp, RxD_t *rxdp)
6508 {
6509         int ip_off;
6510         u8 l2_type = (u8)((rxdp->Control_1 >> 37) & 0x7), ip_len;
6511
6512         if (!(rxdp->Control_1 & RXD_FRAME_PROTO_TCP)) {
6513                 DBG_PRINT(INIT_DBG,"%s: Non-TCP frames not supported for LRO\n",
6514                           __FUNCTION__);
6515                 return -1;
6516         }
6517
6518         /* TODO:
6519          * By default the VLAN field in the MAC is stripped by the card, if this
6520          * feature is turned off in rx_pa_cfg register, then the ip_off field
6521          * has to be shifted by a further 2 bytes
6522          */
6523         switch (l2_type) {
6524                 case 0: /* DIX type */
6525                 case 4: /* DIX type with VLAN */
6526                         ip_off = HEADER_ETHERNET_II_802_3_SIZE;
6527                         break;
6528                 /* LLC, SNAP etc are considered non-mergeable */
6529                 default:
6530                         return -1;
6531         }
6532
6533         *ip = (struct iphdr *)((u8 *)buffer + ip_off);
6534         ip_len = (u8)((*ip)->ihl);
6535         ip_len <<= 2;
6536         *tcp = (struct tcphdr *)((unsigned long)*ip + ip_len);
6537
6538         return 0;
6539 }
6540
6541 static int check_for_socket_match(lro_t *lro, struct iphdr *ip,
6542                                   struct tcphdr *tcp)
6543 {
6544         DBG_PRINT(INFO_DBG,"%s: Been here...\n", __FUNCTION__);
6545         if ((lro->iph->saddr != ip->saddr) || (lro->iph->daddr != ip->daddr) ||
6546            (lro->tcph->source != tcp->source) || (lro->tcph->dest != tcp->dest))
6547                 return -1;
6548         return 0;
6549 }
6550
6551 static inline int get_l4_pyld_length(struct iphdr *ip, struct tcphdr *tcp)
6552 {
6553         return(ntohs(ip->tot_len) - (ip->ihl << 2) - (tcp->doff << 2));
6554 }
6555
6556 static void initiate_new_session(lro_t *lro, u8 *l2h,
6557                      struct iphdr *ip, struct tcphdr *tcp, u32 tcp_pyld_len)
6558 {
6559         DBG_PRINT(INFO_DBG,"%s: Been here...\n", __FUNCTION__);
6560         lro->l2h = l2h;
6561         lro->iph = ip;
6562         lro->tcph = tcp;
6563         lro->tcp_next_seq = tcp_pyld_len + ntohl(tcp->seq);
6564         lro->tcp_ack = ntohl(tcp->ack_seq);
6565         lro->sg_num = 1;
6566         lro->total_len = ntohs(ip->tot_len);
6567         lro->frags_len = 0;
6568         /*
6569          * check if we saw TCP timestamp. Other consistency checks have
6570          * already been done.
6571          */
6572         if (tcp->doff == 8) {
6573                 u32 *ptr;
6574                 ptr = (u32 *)(tcp+1);
6575                 lro->saw_ts = 1;
6576                 lro->cur_tsval = *(ptr+1);
6577                 lro->cur_tsecr = *(ptr+2);
6578         }
6579         lro->in_use = 1;
6580 }
6581
6582 static void update_L3L4_header(nic_t *sp, lro_t *lro)
6583 {
6584         struct iphdr *ip = lro->iph;
6585         struct tcphdr *tcp = lro->tcph;
6586         u16 nchk;
6587         StatInfo_t *statinfo = sp->mac_control.stats_info;
6588         DBG_PRINT(INFO_DBG,"%s: Been here...\n", __FUNCTION__);
6589
6590         /* Update L3 header */
6591         ip->tot_len = htons(lro->total_len);
6592         ip->check = 0;
6593         nchk = ip_fast_csum((u8 *)lro->iph, ip->ihl);
6594         ip->check = nchk;
6595
6596         /* Update L4 header */
6597         tcp->ack_seq = lro->tcp_ack;
6598         tcp->window = lro->window;
6599
6600         /* Update tsecr field if this session has timestamps enabled */
6601         if (lro->saw_ts) {
6602                 u32 *ptr = (u32 *)(tcp + 1);
6603                 *(ptr+2) = lro->cur_tsecr;
6604         }
6605
6606         /* Update counters required for calculation of
6607          * average no. of packets aggregated.
6608          */
6609         statinfo->sw_stat.sum_avg_pkts_aggregated += lro->sg_num;
6610         statinfo->sw_stat.num_aggregations++;
6611 }
6612
6613 static void aggregate_new_rx(lro_t *lro, struct iphdr *ip,
6614                 struct tcphdr *tcp, u32 l4_pyld)
6615 {
6616         DBG_PRINT(INFO_DBG,"%s: Been here...\n", __FUNCTION__);
6617         lro->total_len += l4_pyld;
6618         lro->frags_len += l4_pyld;
6619         lro->tcp_next_seq += l4_pyld;
6620         lro->sg_num++;
6621
6622         /* Update ack seq no. and window ad(from this pkt) in LRO object */
6623         lro->tcp_ack = tcp->ack_seq;
6624         lro->window = tcp->window;
6625
6626         if (lro->saw_ts) {
6627                 u32 *ptr;
6628                 /* Update tsecr and tsval from this packet */
6629                 ptr = (u32 *) (tcp + 1);
6630                 lro->cur_tsval = *(ptr + 1);
6631                 lro->cur_tsecr = *(ptr + 2);
6632         }
6633 }
6634
6635 static int verify_l3_l4_lro_capable(lro_t *l_lro, struct iphdr *ip,
6636                                     struct tcphdr *tcp, u32 tcp_pyld_len)
6637 {
6638         u8 *ptr;
6639
6640         DBG_PRINT(INFO_DBG,"%s: Been here...\n", __FUNCTION__);
6641
6642         if (!tcp_pyld_len) {
6643                 /* Runt frame or a pure ack */
6644                 return -1;
6645         }
6646
6647         if (ip->ihl != 5) /* IP has options */
6648                 return -1;
6649
6650         if (tcp->urg || tcp->psh || tcp->rst || tcp->syn || tcp->fin ||
6651                                                                 !tcp->ack) {
6652                 /*
6653                  * Currently recognize only the ack control word and
6654                  * any other control field being set would result in
6655                  * flushing the LRO session
6656                  */
6657                 return -1;
6658         }
6659
6660         /*
6661          * Allow only one TCP timestamp option. Don't aggregate if
6662          * any other options are detected.
6663          */
6664         if (tcp->doff != 5 && tcp->doff != 8)
6665                 return -1;
6666
6667         if (tcp->doff == 8) {
6668                 ptr = (u8 *)(tcp + 1);
6669                 while (*ptr == TCPOPT_NOP)
6670                         ptr++;
6671                 if (*ptr != TCPOPT_TIMESTAMP || *(ptr+1) != TCPOLEN_TIMESTAMP)
6672                         return -1;
6673
6674                 /* Ensure timestamp value increases monotonically */
6675                 if (l_lro)
6676                         if (l_lro->cur_tsval > *((u32 *)(ptr+2)))
6677                                 return -1;
6678
6679                 /* timestamp echo reply should be non-zero */
6680                 if (*((u32 *)(ptr+6)) == 0)
6681                         return -1;
6682         }
6683
6684         return 0;
6685 }
6686
6687 static int
6688 s2io_club_tcp_session(u8 *buffer, u8 **tcp, u32 *tcp_len, lro_t **lro,
6689                       RxD_t *rxdp, nic_t *sp)
6690 {
6691         struct iphdr *ip;
6692         struct tcphdr *tcph;
6693         int ret = 0, i;
6694
6695         if (!(ret = check_L2_lro_capable(buffer, &ip, (struct tcphdr **)tcp,
6696                                          rxdp))) {
6697                 DBG_PRINT(INFO_DBG,"IP Saddr: %x Daddr: %x\n",
6698                           ip->saddr, ip->daddr);
6699         } else {
6700                 return ret;
6701         }
6702
6703         tcph = (struct tcphdr *)*tcp;
6704         *tcp_len = get_l4_pyld_length(ip, tcph);
6705         for (i=0; i<MAX_LRO_SESSIONS; i++) {
6706                 lro_t *l_lro = &sp->lro0_n[i];
6707                 if (l_lro->in_use) {
6708                         if (check_for_socket_match(l_lro, ip, tcph))
6709                                 continue;
6710                         /* Sock pair matched */
6711                         *lro = l_lro;
6712
6713                         if ((*lro)->tcp_next_seq != ntohl(tcph->seq)) {
6714                                 DBG_PRINT(INFO_DBG, "%s:Out of order. expected "
6715                                           "0x%x, actual 0x%x\n", __FUNCTION__,
6716                                           (*lro)->tcp_next_seq,
6717                                           ntohl(tcph->seq));
6718
6719                                 sp->mac_control.stats_info->
6720                                    sw_stat.outof_sequence_pkts++;
6721                                 ret = 2;
6722                                 break;
6723                         }
6724
6725                         if (!verify_l3_l4_lro_capable(l_lro, ip, tcph,*tcp_len))
6726                                 ret = 1; /* Aggregate */
6727                         else
6728                                 ret = 2; /* Flush both */
6729                         break;
6730                 }
6731         }
6732
6733         if (ret == 0) {
6734                 /* Before searching for available LRO objects,
6735                  * check if the pkt is L3/L4 aggregatable. If not
6736                  * don't create new LRO session. Just send this
6737                  * packet up.
6738                  */
6739                 if (verify_l3_l4_lro_capable(NULL, ip, tcph, *tcp_len)) {
6740                         return 5;
6741                 }
6742
6743                 for (i=0; i<MAX_LRO_SESSIONS; i++) {
6744                         lro_t *l_lro = &sp->lro0_n[i];
6745                         if (!(l_lro->in_use)) {
6746                                 *lro = l_lro;
6747                                 ret = 3; /* Begin anew */
6748                                 break;
6749                         }
6750                 }
6751         }
6752
6753         if (ret == 0) { /* sessions exceeded */
6754                 DBG_PRINT(INFO_DBG,"%s:All LRO sessions already in use\n",
6755                           __FUNCTION__);
6756                 *lro = NULL;
6757                 return ret;
6758         }
6759
6760         switch (ret) {
6761                 case 3:
6762                         initiate_new_session(*lro, buffer, ip, tcph, *tcp_len);
6763                         break;
6764                 case 2:
6765                         update_L3L4_header(sp, *lro);
6766                         break;
6767                 case 1:
6768                         aggregate_new_rx(*lro, ip, tcph, *tcp_len);
6769                         if ((*lro)->sg_num == sp->lro_max_aggr_per_sess) {
6770                                 update_L3L4_header(sp, *lro);
6771                                 ret = 4; /* Flush the LRO */
6772                         }
6773                         break;
6774                 default:
6775                         DBG_PRINT(ERR_DBG,"%s:Dont know, can't say!!\n",
6776                                 __FUNCTION__);
6777                         break;
6778         }
6779
6780         return ret;
6781 }
6782
6783 static void clear_lro_session(lro_t *lro)
6784 {
6785         static u16 lro_struct_size = sizeof(lro_t);
6786
6787         memset(lro, 0, lro_struct_size);
6788 }
6789
6790 static void queue_rx_frame(struct sk_buff *skb)
6791 {
6792         struct net_device *dev = skb->dev;
6793
6794         skb->protocol = eth_type_trans(skb, dev);
6795 #ifdef CONFIG_S2IO_NAPI
6796         netif_receive_skb(skb);
6797 #else
6798         netif_rx(skb);
6799 #endif
6800 }
6801
6802 static void lro_append_pkt(nic_t *sp, lro_t *lro, struct sk_buff *skb,
6803                            u32 tcp_len)
6804 {
6805         struct sk_buff *tmp, *first = lro->parent;
6806
6807         first->len += tcp_len;
6808         first->data_len = lro->frags_len;
6809         skb_pull(skb, (skb->len - tcp_len));
6810         if ((tmp = skb_shinfo(first)->frag_list)) {
6811                 while (tmp->next)
6812                         tmp = tmp->next;
6813                 tmp->next = skb;
6814         }
6815         else
6816                 skb_shinfo(first)->frag_list = skb;
6817         sp->mac_control.stats_info->sw_stat.clubbed_frms_cnt++;
6818         return;
6819 }